Spectrum of Engineering Sciences

MULTI-SOURCE TRAINING ON CROSS-DATASET IN DERMOSCOPIC SKIN CANCER CLASSIFICATION: A 5-FOLD CROSS-VALIDATED STUDY ON HAM10000 AND ISIC 2019 WITH A SOURCE-BALANCED SAMPLER

Muhammad Haroon Ur Rashid — 2026-06-05

Deep learning models for dermoscopic skin cancer classification routinely report accuracies above 90 % on the HAM10000 benchmark, yet their behaviour under realistic cross-domain conditions is rarely measured. This study reports two findings. First, a dual-backbone ConvNeXt–EfficientNet model (66.1 M parameters) trained only on HAM10000 attains a macro-F1 of 0.6905 on the HAM10000 test split but collapses to 0.4301 on the unseen ISIC 2019 archive — a generalisation gap of 26.0 percentage points. Second, a single ConvNeXt-Tiny backbone (27.8 M parameters) trained jointly on HAM10000 and ISIC 2019 with a source-balanced weighted sampler, evaluated under 5-fold lesion-grouped cross-validation with 4-view test-time augmentation and 2 000-resample bootstrap confidence intervals, achieves a pooled macro-F1 of 0.7401 [95 % CI 0.7252, 0.7541] on HAM-test and 0.5976 [95 % CI 0.5817, 0.6128] on ISIC-test. The cross-dataset gap is reduced from 0.260 to 0.142, a 45.3 % reduction, while the backbone shrinks by 58 %. Every per-class F1 improves on both datasets — most dramatically for dermatofibroma (df) on ISIC, which rises from 0.12 to 0.40, and vascular lesions (vasc), which rise from 0.35 to 0.54. The work also surfaces a measurement problem in the recent literature: of ten 2025–2026 studies surveyed, only two report macro-F1 on the 7-class HAM10000 task — and only one of those [8] uses the standard supervised protocol; only one of the ten evaluates true cross-dataset performance. Compared against the single directly comparable cross-dataset benchmark in the recent literature [8], our model achieves a pooled cross-dataset top-1 accuracy of 69.24 % on the ISIC 2019 test set versus their 56.0 % — a 13.2-point improvement — with approximately 2.7× fewer trainable parameters and under a stricter 5-fold cross-validated protocol with bootstrap confidence intervals

LABORATORY EVALUATION OF FLY ASH CENOSPHERE-MODIFIED ASPHALT BINDERS AND ASPHALT MIXTURE THERMAL RESPONSE UNDER CONTROLLED IRRADIATION

Muhammad Safi Ullah — 2026-06-05

Asphalt binders are highly temperature-sensitive materials, and excessive heat accumulation can accelerate softening and reduce pavement service performance in hot climatic regions. This study evaluates the use of fly ash cenospheres (FAC), an industrial by-product with lightweight hollow morphology, as a potential waste-derived modifier for asphalt binders and examines its influence on laboratory-scale thermal response under controlled irradiation. Four binder formulations were prepared: virgin 60/70 penetration-grade binder and binders modified with 5%, 10%, and 15% FAC by binder weight. Conventional binder properties were assessed through penetration, softening point, ductility, flash point, fire point, specific gravity, and rotational viscosity tests. Asphalt mixture slabs prepared with the corresponding binders were then exposed to a controlled irradiation system, and peak temperature and time to peak temperature were recorded. The results showed that FAC modification progressively reduced penetration and ductility while increasing softening point, viscosity, flash point, and fire point, indicating a stiffer and more binder system with improved high-temperature consistency. Under irradiation, the peak temperature decreased from 67.0°C for the control mixture to 59.8°C at 15% FAC, corresponding to a 10.75% reduction. The time to peak temperature increased from 3588 s to 3700 s, indicating delayed heat buildup. These findings suggest that FAC can improve the laboratory thermal response of asphalt mixtures while providing a potential waste-utilization pathway. However, field validation is required before pavement-scale heat-mitigation claims can be made.

ARTIFICIAL INTELLIGENCE IN MANAGEMENT SYSTEMS

Muhammad Adil Shahid — 2026-06-05

The study project in question investigated the implications of the utilization of Artificial Intelligence (AI) technologies on the outcomes of talent management and how it effects employee performance in organizations. This was accomplished by examining the perceptions of AI efficiency, ease of use, training, insights, and general use among employees. In order to carry out the study and compute the descriptive statistics, correlation analysis, and multiple regression, a quantitative survey was conducted with 55 participants. The results of this survey demonstrated that artificial intelligence has the potential to play a significant role in the optimization of talent, with efficiency, ease of use, and AI-generated insights becoming major predictors. Both the general application of artificial intelligence and the training of AI did not result in any statistically significant impacts. This indicates that the value that is generated is not in the adoption of AI systems but rather in the integration of AI systems that are effective, intuitive, and insight based. The findings indicate that strategic integration and quality tool design play a key role in the enhancement of HR processes. Furthermore, the findings suggest that future researchers should increase the size of their samples, take into account additional variables, and employ mixed method approaches in order to get a fundamental comprehension of the subject matter. The results are especially applicable to the organizations that use AI-enabled HR systems, e.g. algorithmic performance dashboards, predictive analytics tools, and machine-learn-based appraisal platforms.

GRAPH-ENHANCED MONOTONIC NEURAL NETWORKS FOR HEALTHCARE OUTCOME REGRESSION

Syed Shaheer Abbas Sherazi — 2026-06-05

Having the ability to estimate healthcare outcomes based on patient data is a significant undertaking in clinical decision-making. Although powerful, conventional regression approaches do not always work to model complexizations of nonlinear connection in medical isotope and deep neural networks, and are often structure-insensitive and uninterpretable. The given paper introduces a novel Graph-Enhanced Monotonic Neural Network (GEMNet) model specifically tailored to work with regression of healthcare outcomes on structured Tabular data. GEMNet provides a trade-off between interpretability and predictive attributes through the embedding of graph neural networks (GNNs) to make predictions of inter-feature connection and by enforcing monotonic implicit constraints based on clinical knowledge. The layers based on the graph convolution are tied to the domain-sensitive domain monotonic activation dominated by the model architecture in such a way that directionally consistency is attained with known risk factors (e.g., age, blood pressure, cholesterol). It has been experimented on a variety of real-world medical datasets (including medical cost prediction and cardiovascular risk estimation) demonstrating that GEMNet tends to perform better than other current regressors, including conventional models, multilayer perceptrons and gradient boosting, in terms of mean squared error (MSE) and R-squared. Better still, the model provides us with interpretable attribution of features and it generalizes better depending on the folds of validation. The results reveal the potential of monotonic graph-based neural models as a scaled-up, clinically-based solution to structured healthcare prediction tasks.

Keywords: Healthcare Outcome Prediction, Graph Neural Networks (GNNs), Monotonic Neural Networks, Tabular Data Regression, Clinical Interpretability, Feature Dependency Modeling, Structured Data Learning, Medical Risk Modeling, Deep Learning in Healthcare, Explainable AI (XAI).

AN INTEGRATED ARTIFICIAL INTELLIGENCE AND INFORMATION TECHNOLOGY FRAMEWORK FOR CLIMATE MODELING AND SUSTAINABILITY DECISION OPTIMIZATION

admin admin — 2026-06-05

The escalating climate crisis demands a radical paradigm shift to extend beyond traditional modeling methods that still suffer due to computational constraints, parameterization gaps as well as the disparity between global modeling and the local decision-making levels. To address these important gaps, the paper demonstrates a transformative Artificial Intelligence and Information Technology framework which combines hybrid physics-AI systems, generative machine learning and policy optimization tools. Using a combination of systematic testing of Neural GCM, dynamical-generative downscaling, physics-constrained neural networks, and analysis of emissions using machine learning, we show new capabilities in climate science never before seen. Physics-AI hybrids reduce errors in precipitation forecasts by 40% but offer 372× computational speeds as well as significantly better representations of extreme precipitation events which previously have been afflicted by systematic drizzle bias. The model of generative diffusion can be used to downscale images at high resolution (greater than 800 samples per hour) and maintain multivariate correlations needed to measure extreme events in compounds- a feature formerly impractical to compute with purely dynamical models. The analysis of the emissions data of 195 countries (1900-2023) that is based on the machine learning reveals carbon intensity of economic activity as the most significant predictive feature (78.0% importance), accompanied by empirically-based national typologies that necessitate differentiated policy interventions, but not the adopted one-size-fits-all measures. More importantly, transfer learning has shown that AI models that are trained under historical conditions can be adapted to new climatic conditions (4×CO2) with only 1% of new training data, which is the root cause of the generalization issue of AI usage in climate science. There is also the operational viability of the framework which is based on proven case studies of monsoon forecasting, infrastructure planning in deep uncertainty, and agricultural decision support systems. This combination of artificial intelligence and information technology with climate science is not just a case of incremental improvement but it is an overhaul of the human ability to comprehend, foresee, and act in response to the accelerating environmental change.

Keywords : Climate modeling, artificial intelligence, hybrid physics-AI systems, generative downscaling, machine learning, extreme events, policy analysis, transfer learning.

AI-MONITORED BIO-CEMENTATION FOR SELF-HEALING IN 3D PRINTED CONCRETE

Ammar Naeem — 2026-06-05

The construction industry worldwide is under unprecedented pressure to create sustainable infrastructure that can autonomously self-repair and real-time health-monitor. Approximately 7% of CO2 emissions are generated in the making of concrete structures, which inevitably develop microcracks that impair durability and structural integrity. Traditional methods of cracking repair, such as grouting, epoxy-injection, and manual repairing, are time-consuming, expensive, and harmful to the environment. To overcome these drawbacks, bio-cementation-based self-healing concrete with the mechanism of microbially induced calcium carbonate precipitation (MICP) has become a novel sustainable solution. This bio-precipitation technique using bacterial strains such as Bacillus cereus, Bacillus megaterium, and Bacillus licheniformis has the capacity of self-healing of concrete up to 0.97 mm with 14–32% mechanical strength improvement. Not only does this biological mineralization help to restore structural integrity, it also has a massive impact on increasing of the durability and resistance to chemical attack. However, to fully realize the potential of MICP future challenges such as viability of bacteria in alkaline conditions, nutrient optimization and scaling should be overcome. Emerging technologies such as three-dimensional concrete printing (3DCP), artificial intelligence for crack detection, and smart sensing offer novel possibilities for conducting real-time structural health monitoring (SHM), an important factor in predictive management. This review aims to summarize existing knowledge about bio-cement 3DCP, including: MICP mechanisms and bacterial encapsulation strategies; bio-cement 3DCP materials design and sustainability; AI monitored SHM systems and integrated system performance and future commercialization. This broad review of 20+ primary sources sets a critical framework between microbial self-healing, digital manufacturing and intelligent monitoring. As the evidence shows, bio-cemented 3D-printed concrete can have a compressive strength of more than 50 MPa and it can offer up to 48% reduction of Embodied Carbon as well as ensuring the capacity of autonomous damage detection and repair.

THE EVOLUTION OF INFORMATION TECHNOLOGY IN THE AGE OF ARTIFICIAL INTELLIGENCE: OPPORTUNITIES AND STRATEGIC CONSEQUENCES

*Aziz Khan — 2026-06-05

Background: The fast developing concept of the Artificial Intelligence (AI) has radically altered the current Information Technology (IT) systems, organizational procedures, and digital innovation tactics. Machine learning, intelligent automation, predictive analytics, and cloud-based systems are examples of AI-powered technologies that are redefining operational efficiency, cybersecurity, strategic planning, and organizational competitiveness. Regardless of these developments, organizations are still grappling with strategic issues regarding ethical issues, workforce flexibility, costs of implementation, and governance constraints. Objective: This paper set out to discuss the transformational effects of Artificial Intelligence on Information Technology systems, determine the opportunities afforded by the integration of AI, consider the strategic implications and organizational issues related to the use of AI, and discuss how organizations will be prepared to work in AI-driven technological contexts in the future. Methodology: The quantitative method of research was embraced by a structured questionnaire that was carried out to 280 respondents who were related to the Information Technology industry as IT professionals, academics, managers, researchers and entrepreneurs. The data were collected using a five-point Likert scale measure, which comprised of 23 items, categorized under four broad constructs AI Technological Transformation, AI Opportunities in IT, Strategic Consequences and Challenges, and Strategic Readiness and Future Outlook. The data collected was analyzed using SPSS that included descriptive statistics, Cronbach Alpha reliability test and chi-square analysis. Results: The results showed good overall instrument reliability with Cronbachs Alpha at 0.906. The average score was 4.08, which showed that the respondents had a very positive attitude towards Artificial Intelligence. The highest mean score (M = 4.26) was the construct AI Opportunities in IT, which means that there was the highest agreement regarding the use of AI in automation, increased cybersecurity, innovation, and competitive advantage. On the same note, the level of agreement in AI Technological Transformation was found to be high (M = 4.18), which proves that AI plays an important role in changing digital infrastructures and organizational processes. However, the strategic implications and organizational concerns were moderate-to-high among the respondents (M = 3.74) particularly in the context of threats to privacy, ethical concerns, displacement of the workforce and implementation cost. In addition, it was observed that the organizational preparedness in the future implementation of AI was high (M = 4.12), and it revolved around the role of AI governance structures, investments in innovation, and AI-human relationships. Conclusion: The research concludes that Artificial Intelligence is now a revolutionary element in the contemporary Information Technology set-ups by improving innovation, operational efficiency, and strategic competitiveness. Nevertheless, companies should implement responsible governing approaches, enhance cybersecurity measures, and focus on reskilling their workforce to better deal with the strategic challenges AI adoption implies. To ensure sustainable AI implementation, a moderate path will need to be taken, which involves technological progress and ethical accountability and control.

Keywords : Artificial Intelligence, Information Technology, Digital Transformation, Strategic Management, Cybersecurity, AI Governance, Organizational Readiness, Intelligent Automation

AI-DRIVEN OPTIMIZATION OF PEROVSKITE SOLAR CELLS FOR SUSTAINABLE ENERGY DEVELOPMENT IN PAKISTAN

Zain Nawazish — 2026-06-06

The findings confirm that AI-integrated PSC systems can substantially contribute to improving renewable energy generation efficiency, reducing dependency on fossil fuels, and supporting Pakistan’s long-term energy security and sustainability goals.

The transition toward sustainable and low-carbon energy systems has intensified global research into high-efficiency photovoltaic technologies. Perovskite solar cells (PSCs) have emerged as a promising alternative to conventional silicon-based photovoltaics due to their high power conversion efficiency, low-cost fabrication potential, and tunable optoelectronic properties. However, challenges such as environmental instability, thermal degradation, ion migration, and limited long-term operational reliability continue to hinder large-scale commercialization. Artificial Intelligence (AI), including machine learning, deep learning, and predictive analytics, has recently demonstrated strong potential in accelerating materials discovery, optimizing device architectures, and improving photovoltaic performance prediction. This study investigates the role of AI-driven optimization in enhancing the efficiency, stability, and operational performance of PSCs, with a specific focus on sustainable energy development in Pakistan. A quantitative explanatory research design was employed using data from 350 professionals working in renewable energy, artificial intelligence, and photovoltaic-related fields. Data were analyzed using Structural Equation Modeling (SEM) and regression techniques. The results revealed that AI capability significantly enhances PSC optimization, which in turn strongly influences sustainable energy development. The model explained 72.4% of the variance in sustainability outcomes, indicating strong predictive validity.

PROJECT MANAGEMENT PRACTICES AS ENABLERS OF AI-DRIVEN DIGITAL TRANSFORMATION : AN EMPIRICAL INVESTIGATION OF THEIR IMPACT ON ORGANISATIONAL PERFORMANCE

Umar Farooq — 2026-06-06

Artificial intelligence is increasingly central to digital transformation because it enables predictive decision-making, intelligent automation, process redesign, and customer-focused service innovation. Yet AI-enabled transformation frequently fails to deliver expected organisational value when implementation is treated as a technical deployment rather than a managed organisational change initiative. This report examines project management practices as enablers of AI-driven digital transformation and analyses their impact on organisational performance. Drawing on digital transformation literature, agile project management research, dynamic capabilities theory, stakeholder theory, and AI governance scholarship, the report proposes an integrated framework linking agile and hybrid delivery, stakeholder engagement, risk governance, change management, data and resource integration, and benefits realisation to AI transformation success. A mixed-method empirical design is presented, supported by an illustrative dataset of 286 project and digital transformation professionals. The illustrative findings indicate that agile delivery, stakeholder engagement, and risk governance are strongly associated with AI transformation success, while benefits realisation and post-implementation monitoring are particularly important for translating AI deployment into measurable organisational performance. The report includes a full conceptual framework, AI transformation lifecycle, research design model, tables of constructs, hypotheses, illustrative results, and practical recommendations for organisations seeking to improve AI adoption outcomes.

HYBRID COGNITIVE AI FRAMEWORKS FOR INTELLIGENT ENGINEERING SYSTEMS: INTEGRATING MACHINE LEARNING AND SYMBOLIC REASONING

Rehan Ali Khan — 2026-06-06

Artificial intelligence has become a transformative technology in intelligent engineering systems, creating opportunities for enhanced automation, decision-making, and operational efficiency. This study investigated the role of Hybrid Cognitive AI Frameworks in improving Intelligent Engineering System Performance through the integration of Machine Learning and Symbolic Reasoning. A quantitative research design was employed, and data were collected from a sample of 320 engineering professionals, AI specialists, software engineers, and technology practitioners. The study examined the relationships among Machine Learning, Symbolic Reasoning, Hybrid Cognitive AI Frameworks, and Intelligent Engineering System Performance using descriptive statistical techniques. The findings indicated strong positive perceptions regarding all study variables. Machine Learning achieved a mean score of 4.31 with a standard deviation of 0.58, Symbolic Reasoning recorded a mean score of 4.24 with a standard deviation of 0.61, Hybrid Cognitive AI Frameworks achieved a mean score of 4.36 with a standard deviation of 0.55, and Intelligent Engineering System Performance recorded the highest mean score of 4.41 with a standard deviation of 0.53. The 84.4% respondents agreed that the integration of Machine Learning and Symbolic Reasoning enhanced engineering intelligence and system effectiveness. The findings suggested that hybrid cognitive AI approaches improved explainability, adaptability, reliability, and operational efficiency within engineering environments. The study concluded that integrating learning-based and reasoning-based AI paradigms supported the development of intelligent, transparent, and trustworthy engineering systems capable of addressing complex technological challenges.

Keywords : Artificial Intelligence, Hybrid Cognitive AI Frameworks, Intelligent Engineering Systems, Machine Learning, Neuro-Symbolic AI, Symbolic Reasoning.

https://doi.org/10.5281/zenodo.20570286

NANO-HYBRID: A LIGHTWEIGHT INCEPTIONNEXT-ATTENTION NETWORK FOR EFFICIENT LUNG CANCER DIAGNOSTICS

Hazik Jaffri — 2026-06-06

Lung cancer remains a leading cause of cancer-related mortality, necessitating diagnostic tools that are both accurate and computationally efficient for widespread deployment. While recent hybrid Deep Learning models have achieved high classification performance, they typically rely on heavy architectures (>18 million parameters) and extensive pre-training, limiting their applicability on resource-constrained edge devices. This study proposes a Nano-Hybrid architecture that integrates lightweight InceptionNeXt convolutions with global attention mechanisms, designed to be trained entirely from scratch. We evaluated the model on two diverse datasets: the IQ-OTH/NCCD (3-class) and a multi-class Chest CT dataset (4-class). Despite containing ~89% fewer parameters (2.03M) than comparable state-of-the-art baseline models, our approach achieved 95.41% accuracy on the IQ dataset, demonstrating that massive capacity is not strictly required for high-performance diagnostics. On the challenging multi-class Chest CT dataset, the model achieved 86.11% accuracy, with a notable 1.00 AUC (Area Under Curve) for normal cases, ensuring zero false positives in healthy screenings. Explainability analysis using Grad-CAM further validates that the model correctly prioritizes pulmonary nodule structures over background artifacts.

A SYSTEMATIC LITERATURE REVIEW OF LSTM NETWORKS FOR BITCOIN PRICE FORECASTING PERFORMANCE

Shabbir Ahmad — 2026-06-06

Bitcoin price forecasting remains a significant challenge in financial analytics due to the highly volatile and nonlinear nature of cryptocurrency markets. Traditional forecasting techniques, such as Autoregressive Integrated Moving Average (ARIMA) and Linear Regression, often struggle to capture the complex temporal relationships present in Bitcoin price movements. In recent years, deep learning approaches, particularly Long Short-Term Memory (LSTM) networks, have gained considerable attention because of their ability to model sequential data and learn long-term dependencies. This study presents a Systematic Literature Review (SLR) of the performance of pure LSTM networks in Bitcoin price forecasting. Following the PRISMA framework, 60 peer-reviewed studies published between 2020 and 2026 were systematically identified and analyzed from major academic databases, including Google Scholar, IEEE Xplore, ScienceDirect, and SpringerLink. The review evaluates forecasting accuracy, methodological consistency, interpretability, and the effectiveness of standardized OHLCV (Open, High, Low, Close, and Volume) data in prediction tasks. The findings indicate that pure LSTM models generally outperform traditional econometric methods in highly volatile market conditions due to their gated memory architecture, which effectively captures long-term temporal patterns. The study highlights the potential of LSTM as a reliable and interpretable forecasting approach and provides a benchmark framework for future research in cryptocurrency forecasting and artificial intelligence-driven financial analytics.

INTERPRETABLE DEEP LEARNING MODELS FOR CLASSIFICATION OF BRAIN TUMORS VIA MRI

Ilya Haider — 2026-06-06

Brain tumors are super serious neurological issues, and getting them diagnosed quickly and right is key for better outcomes and effective treatment plans. Doctors use Magnetic Resonance Imaging (MRI) a lot because it does the best job of showing soft tissues and giving detailed views of the brain. Recently, tools like Convolutional Neural Networks (CNNs) in deep learning have gotten really good at classifying these tumors automatically. Yet, there’s a catch – these models are like black boxes; no one can see how they make decisions. This makes doctors and other health pros wary about using them. Our study aims to tackle this by coming up with an Explainable Deep Learning (XDL) framework. It lets us classify brain tumors accurately from MRI scans while also making it clear how those decisions are reached. The proposed method uses a deep convolutional neural network, trained on processed MRI images, to classify brain tumors into types like glioma, meningioma, and pituitary tumors. To boost transparency and clinician trust, they added explainability techniques, including Grad-CAM, LIME, and attention visualization. These methods show which parts of an image influenced the model's decision, helping radiologists see why the system thinks a tumor is one type over another tests show that this model performs really well in terms of accuracy, precision, recall, F1-score, and AUC. It does this while offering clear visual explanations too. This proves that explainable AI can help bridge the gap between tech and healthcare decisions. By doing so, it makes AI models more reliable, transparent, and trustworthy for doctors. The work fits into the bigger picture of making medical AI trustworthy and supports radiologists in accurately and clearly diagnosing brain tumors.

AHP-BASED WEIGHTING APPROACH FOR RISK ASSESSMENT AND PRIORITIZATION IN PAKISTAN’S COAL SUPPLY CHAIN

Muhammad Asad Ullah — 2026-06-06

This study investigates the risks and mitigation strategies associated with Pakistan's coal supply chain, focusing on enhancing its resilience and sustainability. The research uses a mixed-methods approach, combining qualitative and quantitative data collection techniques, such as conducting semi-structured interviews and a structured survey questionnaire, involving 120 stakeholders. Data was analysed quantitatively using SPSS. Ranking and prioritizing of identified risks were done using Analytical Hierarchy Process (AHP) based weighting approach. High priority weighting shows that the quality factors (0.134-0.154) and the time factors (0.104-0.135) are more important. The most important individual risk factors are total moisture (0.154) and lead time (0.135). Furthermore, the study shows that the weak infrastructure, geopolitical instability, regulatory issues, and environmental issues like pollution and carbon emissions are the main concerns in the coal supply chain in Pakistan. Technology improvements, including energy-efficient machines and mining technology, were seen as important contributors to the improvements in the supply chain. The findings of the study corroborate previous studies and bring to the fore issues that are specific to Pakistan including the effect of policy fluctuations and geopolitical tensions in the region. The study highlights the importance of strategic investments in infrastructure, sustainable practices, technological innovation, and collaboration among stakeholders to improve the resilience of the coal supply chain and support the nation's energy security and economic development.

COMPARATIVE ANALYSIS OF MACHINE LEARNING TECHNIQUES FOR NETWORK INTRUSION DETECTION IN CYBER SECURITY WITH A DIVERSE METRIC-BASED PERFORMANCE ASSESSMENT

Farhan Tariq — 2026-06-06

n modern communication and networking, the safe and reliable transfer of data is a necessity of time because the number of intruder attacks on computer networks aims to gain access to crucial information. To protect the network data from any malicious attack, the network intrusion detection systems (NIDSs) play the most critical role. It analyzes the data pattern and secures the network from any attack. This pattern analysis is not possible manually due to the large scale of data; however, machine learning (ML) is a powerful technique to analyze the large scale of data patterns and detect any malicious threats. In this work, we integrated ML with NIDS to analyze and monitor the networking data. We have applied six supervised ML techniques, which include Random, Hoeffding, and Decision Tree, Averaged One-Dependence Estimators, Instance-based KNN, and Naive Bayes, during the experiment and also considered six performance assessment criteria, which include accuracy, precision, true and false positive rates, Matthew correlation coefficient, and receiver operating characteristic area for the three different datasets. The Pareto principle is considered for the training and testing data. According to the results, A1DE is the best model among the applied techniques; it identifies patterns in the data with 99.9964% accuracy, which establishes a foundation for further research. The researchers use these findings as a starting point for determining which cyber-related attributes should be prioritized to create the most effective and successful NIDS.

ENHANCING TEACHING AND LEARNING THROUGH GENERATIVE ARTIFICIAL INTELLIGENCE: BENEFITS, CHALLENGES, AND ETHICAL CONSIDERATIONS

Awais Maqsood — 2026-06-06

Generative AI has significantly transformed learning, teaching, and academic support since the release of OpenAI’s ChatGPT in 2022. It is changing how students learn, how teachers teach and how support is provided in schools and universities. Generative AI helps students collaborate, improve creativity, solve problems, and learn independently, especially in subjects like mathematics, physics, and coding. It is also used to create quizzes, summaries, and personalized learning materials that make complex topics easier to understand. Research shows that these tools can improve academic performance, motivation, confidence, and creativity among students. However, concerns remain about cheating, plagiarism, privacy, and unequal access to technology. Because of these challenges, Generative AI should not replace teachers but should be used as a supportive educational tool. With proper teacher training, student awareness, and ethical use, Generative AI has strong potential to improve education and make learning more effective and accessible for everyone.

AN ADVANCED METHOD FOR CHANNEL FADING PARAMETER ESTIMATION BASED ON THE GENERALIZED GAMMA DISTRIBUTION

Muhammad Ilyas — 2026-06-06

In this manuscript, we present statistical characterization of the fading of the radio channels is of crucial importance to the planning and testing of today's mobile communication networks. The extended gamma (Stacy) variation is a very flexible fading model and includes a number of common distributions, e.g. Rayleigh, Gamma, Weibull, and Nakagami-m. Although it is a general model, it has not been easy to obtain reliable estimates of its key parameters, especially in cases where limited measurement data are available. In this experiment, a novel Psi-inverse (PI) parameter estimation method for the generalized gamma fading model is suggested and its performance is evaluated with respect to a maximum likelihood estimator. The proposed method is based on the use of digamma-based transformations to achieve better numerical stability and numerical accuracy. Its performance is systematically compared to conventional estimation methods, e.g. method-of-moments and skewness-logarithmic estimator. Extensive Monte Carlo simulations are carried out in many different fading scenarios and sample sizes typical of real-world wireless scenarios. The results demonstrate the PI estimator is always superior to the existing methods especially in the regimes of small or moderate samples, where the conventional ones tend to be biased and unstable. While the maximum likelihood estimator is fine for large data sets, the estimator is not as reliable if the availability of the data is limited. The major achievement of this work is the introduction of a powerful parameter estimation method with computational efficiency, which yields a great increase in the estimation accuracy under actual operational conditions. The proposed method is well suited for practical applications for wireless channel modeling, system simulation and analysis of performance of communication systems operated in complex fading environment.

Keywords : Fading radio signals, distribution, the Stacy distribution, Gamma distribution, Erlang distribution, Chi-squared distribution, Nakagami distribution, size biased distributions, ML estimators.

ASSESSING THE EFFECTIVENESS OF DDOS MITIGATION STRATEGIES THROUGH NETWORK EMULATION

Abdul Qadir — 2026-06-05

Research domain or Background The Distributed Denial of Service (DDoS) attacks pose among the most persistent and increasingly threatening problems in the modern age of network infrastructure due to their capability to exhaust the bandwidth, processing capabilities, connection tables, and memory of the targeted system. Research Problem Efficiently emulating such attack scenarios under economically feasible circumstances and in a controllable manner is indeed difficult yet highly necessary for academic and commercial security assessment purposes. Research Objective In this paper, we conduct an organized and well-designed emulation experiment involving a simulation of DDoS attacks (specifically ICMP, UDP, and TCP SYN floods) on a real-world network configuration consisting of Cisco routers and switches, a web server, legitimate client machines, and a Kali Linux machine acting as the attacking agent. Research Design/Methodology Five layers of mitigation techniques have been used and tested; these included VLAN segmentation, access control list (ACL), port security, rate limit, and Quality of Service (QoS). Research Findings The experimental data shows that the application of all these techniques reduces the influence of a DDoS attack on legitimate traffic but also does not affect their performance. Research Limitations Statistical analysis proves that GNS3 is efficient in testing DDoS attacks at medium to lower rates because the maximum attack traffic was set at 10,000 packets per second and 100 megabits bandwidth. This research highlights important issues associated with scalability, diversity, and effectiveness of simulation, attack, and protection mechanisms, and suggests research directions including ML attack detection and SDN techniques.

Keywords : DDoS, Network Emulation, GNS3, ICMP Flood, TCP SYN Flood, UDP Flood, ACL, VLAN, QoS, Rate Limiting, Port Security, Kali Linux, hping3, Network Security, Botnet Simulation, Traffic Analysis.

AI AND INTELLIGENT PROJECT MANAGEMENT

Azhar Mehmood* — 2026-06-07

Artificial Intelligence (AI) is transforming project management with enhanced project planning, project execution, and risk management. AI further streamlines the decision-making process. This research examined the state of AI in project management using a systematic literature review (SLR) based on the PRISMA 2020 guidelines. Using the PRISMA methods, 120 peer-reviewed articles on AI and project management published between 2018 and 2025 were collected and analyzed. Five databases were searched: Scopus, Web of Science, Science Direct, IEEE Xplore, and Google Scholar. The applications, advantages, and trends of AI in project management were the focus of these articles. The outcomes showed more research was conducted in the review period, thus showing more project-based organizations were adopting AI. The most cited forms of AI were Machine Learning and Predictive Analytics. These forms of AI were applied to project management functions including, but not limited to, planning, scheduling, risk management, decision-making, project management, and performance. AI was shown in all cited articles to enhance decision-making, improve management of project risks, improve project efficiency, improve management of project resources, and improve project time management. AI in combination with digital transformation was shown to help organizations move from a reactive approach to project management and planning to a proactive approach. Data management, AI algorithm transparency, research on AI ethics, and AI skills are still barriers to the widespread adoption of AI. AI is proving to be a key competitive advantage to organizations that wish to use project management to improve performance. Further studies need to concentrate on explainable AI, applications of generative AI, human-AI collaboration, and governance frameworks that facilitate the functional and responsible use of AI within project settings.

A COMPUTATIONAL MATHEMATICAL FRAMEWORK FOR HIGH-DIMENSIONAL ENGINEERING DATA ANALYSIS USING ADVANCED LINEAR ALGEBRA, MATRIX FACTORIZATION AND OPTIMIZATION TECHNIQUES

Muhammad Umair Aslam — 2026-06-08

High-dimensional engineering datasets are increasingly generated from smart sensors, simulation platforms, industrial monitoring systems, communication networks, and intelligent control environments. However, the large number of variables, nonlinear relationships, redundant features, and computational complexity often reduce the efficiency and accuracy of conventional data analysis methods. This study presents a mathematical framework for high-dimensional engineering data analysis using advanced linear algebra and optimization techniques. The proposed framework integrates matrix decomposition, vector space transformation, dimensionality reduction, eigenvalue-based feature representation, convex optimization, gradient-based learning, and regularization methods to improve data processing, predictive modeling, and machine learning performance in engineering applications. The methodology focuses on transforming complex engineering datasets into optimized mathematical representations by applying Principal Component Analysis, Singular Value Decomposition, least-squares optimization, and regularized regression techniques. These methods help reduce noise, remove irrelevant features, enhance computational speed, and improve model interpretability. The optimized feature space is then used with machine learning models such as Support Vector Machine, Random Forest, and Neural Network classifiers for predictive analysis and decision support. Experimental results demonstrate that the proposed mathematical framework significantly improves model performance compared with traditional feature-processing methods. The dimensionality of the dataset was reduced by 42.6%, while preserving 96.8% of the original data variance. The proposed framework achieved an overall prediction accuracy of 97.3%, precision of 96.5%, recall of 95.9%, and F1-score of 96.2%. In addition, computational training time was reduced by 31.4%, and mean squared error decreased from 0.084 to 0.031 after applying optimization-based feature transformation. The results confirm that advanced linear algebra and mathematical optimization techniques provide a strong foundation for high-dimensional engineering data analysis. Overall, this research highlights the importance of mathematical modeling in improving machine learning efficiency, predictive accuracy, and intelligent decision support for modern engineering systems. The proposed framework can be applied in areas such as smart manufacturing, structural health monitoring, electrical systems, robotics, and engineering design optimization

A SMART HELMET TO DETECT ANOMALIES OF ITS USERS AND ENVIRONMENT

Aqsa Khursheed — 2026-06-08

The Smart Helmet is an intelligent device that can tell if a person is wearing a helmet and if they are driving with non-alcoholic breath. In this case, we have a transmitter on the helmet and a receiver on the bike. A switch ensures that the helmet is always on the user’s head. The ON state of the switch guarantees that the helmet is correctly placed. Alcohol sensors are installed near the rider's mouth; if any of these conditions are not met, then the engine can't start. If the rider is involved in an accident and the helmet is thrown to the ground, then the alcohol sensor detects this and activates the GSM Module, which automatically contacts a family member. It is our main goal to make it easier for motorcycle riders to see on the road. Ultrasonic sensors and a vibrator motor in the new system can measure the necessary distances between passing motorcycles and the vehicle in the rear. The system will alert the rider through the vibrator motor, LEDs, and buzzer that are installed on their helmet as a warning to them about the range of insecurity that the ultrasonic sensor detects. Arduino UNO was used as the system's primary processing unit to manage all the system's networking elements. Arduino UNO put in front of the rider and displays the distance detected by the ultrasonic sensor using OLED displays. Data transmitted by the ultrasonic sensor will be wirelessly transferred to the helmet node, which serves as a reception unit, using the wireless transceiver module.

MASF: MASKED ASYMMETRIC SPECTRAL FLOW FOR UNSUPERVISED INDUSTRIAL ANOMALY DETECTION

Maheem Khowaja — 2026-06-08

An important problem in the industrial quality control application is that of unsupervised anomaly detection, in which only defect free training images are available. All the currently available state-of-the-art techniques such as PaDiM, PatchCore, RD++, CFA, ISSTAD, and ADTR are in the spatial domain and miss the part of the information in the frequency domain of CNN feature maps that has anomaly-discriminative information. We introduce the first framework to leverage frequency domain at the CNN feature level, called MASF (Masked Asymmetric Spectral Flow). MASF introduces five novel components: (1) a Spectral Frequency Decomposition Module (SDM) based on 2D FFT on intermediate feature maps; (2) Asymmetric Masked Feature Distillation (AMFD) using dual spatial-frequency domain masking and Spectral-Spatial Cross-Attention (SSCA) fusion; (3) a Spectral-Anchored Memory Bank (SAMB) for rotation-robust prototype retrieval; (4) Uncertainty-Gated Hierarchical Score Fusion (UGHF) with learnable per-scale precision weights; and (5) Test-Time Spectral Augmentation (TTSA) by FFT phase perturbation. On the Bottle category, evaluated on the MVTec Anomaly Detection benchmark, MASF gets Image-AUROC = 99.9999, Pixel-AUROC = 0.9853, PRO = 0.9461, and AP = 99.9999. In 11 stable training categories, MASF reaches the performance of mean Image-AUROC = 0.8008 and mean Pixel-AUROC = 0.9380 with just 15 training epochs. The design of MASF is directly motivated by the results of FFT spectral analysis which shows that the frequency signature of the normal image is different from that of the anomalous image, as shown by industrial images.

OXYGEN EVOLUTION REACTION BY USING PHOTOANODIC TA3N5 FOR WATER SPLITTING PROCESS BY DIFFERENT SURFACE MODIFICATION: A REVIEW

Toaqeer Salman — 2026-06-08

Fossil energy is a widely used energy source these days, but because of the fossil usage; many complications also arise. In response, a global shift toward sustainable and renewable energy sources has amplified interest in photoelectrochemical (PEC) water splitting as a viable route for clean hydrogen production. Water splitting, which involves the decomposition of water into hydrogen and oxygen, depends critically on the development of efficient and stable semiconductor photoanodes. For this reason, many semiconductors are used; but titanium nitride (Ta₃N₅) semiconductor has great importance because of the low-over potential, better band structure, lesser charge transfer resistance (Rct), decreased solution resistance (Rs), maximum current density and abundance. However, the practical application of Ta₃N₅ is limited by poor charge mobility, surface instability, and rapid electron-hole recombination. To overcome these limitations, significant research has been carried out to prepare the nanocomposites of Ta₃N₅ such as nanofibers, nanofilms, micro sheets, dum bell-like nanostructures, and nanoflowers. These varied morphologies not only enhance visible-light harvesting and charge separation but also lower overpotential and suppress recombination losses, thus improving the overall efficiency of PEC water splitting. Furthermore, a variety of synthetic methodologies including hydrothermal, sol-gel, electrospinning, electrochemical, precipitation, and chemical reduction techniques helped in achieving uniform doping, nanoscale control, and enhanced structural stability. In conclusion, Ta₃N₅ is of significant interest in semiconductor research for water splitting applications. However, future research must focus on improving long-term operational stability, enhancing charge transport across interfaces, and integrating Ta₃N₅ into tandem PEC cells or hybrid solar fuel systems.

QUANTUM CONFINED WATER IN POLYMERIC NANOCHANNELS: PROTON TRANSPORT AND ELECTROCHEMICAL IMPLICATIONS FOR GREEN HYDROGEN ELECTROLYSIS

Sumera Zaib — 2026-06-08

The confining of water inside sub-nanometer to nanometer-sized polymeric channels introduces novel structural and dynamical characteristics that significantly impact proton transport, with fundamental implications for the development of green hydrogen electrolysis. Within such confinement, water molecules tend to assemble into quasi-one-dimensional chains or ordered hydrogen-bond networks, facilitating proton conduction processes that are deviant from bulk conditions. Molecular simulations and neutron scattering measurements have shown that nuclear quantum effects (NQEs) are instrumental in decreasing the free-energy barrier for proton transfer, typically resulting in nearly barrierless conduction regimes. They result from proton delocalization, zero-point energy contributions, and stabilization of Grotthuss-like hopping mechanisms along aligned water chains. Carbon nanotube and hydrophobic nanochannel studies indicate that confined systems can increase proton mobility by several orders of magnitude over bulk water, a characteristic that can be engineered in polymeric electrolytes like Nafion and customized nanocomposites. In addition, modified hydrogen-bond fluctuations under confinement have been linked to increased ionic conductivity and lower activation energy for electrochemical processes. These results offer a basic template for the engineering of future-generation polymer electrolyte membranes, in which quantum-confined water channels can be tapped to enhance the efficiency and longevities of green hydrogen electrolyzers

TECHNICAL AND NON-TECHNICAL LOSS ANALYSIS IN PAKISTANI DISTRIBUTION COMPANIES (DISCOS): CAUSES, ECONOMIC IMPACT AND MITIGATION STRATEGIES

Engr. Muhammad Bilal Ahmad — 2026-06-08

Pakistan’s power sector faces a perpetual socio-economic crisis characterized by escalating circular debt, highly volatile operational inefficiencies and large-scale financial imbalances. At the heart of this structural collapse lie severe transmission and distribution (T&D) losses within the Power Distribution Companies (DISCOs). This comprehensive research paper provides an extensive diagnostic evaluation of Technical Losses (TL) and Non-Technical Losses (NTL) across six prominent DISCOs: IESCO, LESCO, K-Electric, PESCO, HESCO and SEPCO. Technical losses, originating from line resistance, aging transformers, overloaded feeders and unoptimized high-voltage transmission layouts, are systematically distinguished from non-technical losses, which comprise direct power theft via illegal hooking (kundas), advanced meter tampering, systemic billing inaccuracies and abysmal revenue collection efficiencies. Utilizing multi-year empirical datasets spanning from 2018 to 2025 derived from the National Electric Power Regulatory Authority (NEPRA), the Ministry of Energy and individual corporate distribution audits, this study conducts statistical trend mapping, comparative performance evaluation and rigorous economic impact analysis. The empirical evidence reveals a dramatic polarization: while IESCO and LESCO demonstrate robust operational performance with T&D losses stabilizing near NEPRA-allowed limits (8.2% and 11.4% respectively), peripheral DISCOs such as PESCO, HESCO and SEPCO suffer from catastrophic, unmitigated losses exceeding 37%, driven primarily by pervasive commercial theft and deeply institutionalized billing recovery inefficiencies. Economically, these losses directly exacerbate the national circular debt—which has reached an alarming PKR 2.48 trillion by fiscal year 2025—choking public liquidity and severely constraining macroeconomic growth. To reverse this structural hemorrhage, this study proposes a comprehensive, multi-layered technological framework anchored on Advanced Metering Infrastructure (AMI), automated distribution transformer energy balancing, artificial intelligence-driven data mining for predictive fraud detection and robust legal-institutional reforms. This integrated blueprint offers a realistic path toward financial stability, system reliability and sustainable energy governance within Pakistan's power network

PREDICTING THE PRICE OF AUCTION CARS WITH MACHINE LEARNING ALGORITHMS

Muhammad Nadeem — 2026-06-08

The problem of the automotive auction market to estimate the price of the cars accurately becomes critical as the number of features and interaction between these features grows and the conditions are also not standardized. In the present study, three machine learning algorithms—Linear Regression, Random Forest Regression, and an Extreme Gradient Boosting (XGBoost)—are compared using a unique dataset that was developed by integrating past data from car auctions to predict the prices of cars at auctions. Specific data features for the domain were also added, such as make, model, manufacturing year, engine, mileage, exterior color, chassis code, package trim and standardized auction condition grades (1.0 through 5.0). All missing value imputations, label encoding, Z-scores normalization, and more complex feature engineering methods, such as Vehicle Age, Mileage Intensity, Luxury Brand Mapping, and Make-Model Interaction terms have been performed prior to the processing phase. Performance of models was measured by Mean Absolute Error (MAE), Root Mean Squared Error (RMSE) and R-squared (R²) measures. Experimental results show that the accuracy of prediction of XGBoost is observed to be the highest with R² = 96.68%, MAE = 1,403, and RMSE = 1,975 which is higher than the accuracy of Random Forest (R² = 0.9527) and Linear Regression (R² = 0.8321). The results confirm the previous findings that ensemble-based gradient boosting methods improve considerably against linear models abilities when the price estimation is a dedicated task of an auction domain, particularly when feature engineering is employed to enhance the abilities

ANALYZING MACHINE LEARNING TECHNIQUES FOR DETECTION OF NEURODEGENERATIVE DISEASES

Unzila Nasir — 2026-06-08

Neurodegenerative disorders belong to the list of the major causes of the global burden of disease, and scientists urgently require creation of the new methodological instruments to assist in the diagnosis of the early pathological change. Recent machine learning (ML) models observe the importance of appropriate pre- processing of input of other nature. Several research studies have found out that researchers employ multimodal representations in order to stimulate a substantial improvement in predictive performance. The second trend of this nature in long-term healthcare area is the extension of the concept, care to his own. The availability of the possibility to recognize potential indicators depends on the developed machine learning processes since the methods are able to accommodate image, electrophysiological and multi- modals. The recent machine learning machines have highlighted the importance of pre-processing. Several studies have shown that researchers consider multi-modes to be instrumental when combined. In the recent times, neuroscience computational frameworks have shown the importance of early extraction of the biomarkers. The present review is an amalgamation of the existing methodological developments, suggests the implementation of specific diseases, conflicts the behavior of models, and chances in optimizing. The current modelling schemes focus on the importance of optimal pipelines. The current ideas of the computational neuroscience have stimulated the attempts to find and isolate biomarkers during the early stages. This development can be seen in the society in general.

INTEGRATION OF MACHINE LEARNING WITH BLOCKCHAIN FOR HEALTHCARE SYSTEM

Salma Rasool — 2026-06-08

The developments that took place in the fields of Blockchain Technology, Internet of Things (IoT), and Machine Learning (ML) offer promising and exciting developments within the paradigm shift that took place within various sectors, especially within the health industry. The intersection of these three will address issues on data security and privacy within the context of real-time decision-making. Within this research study, the intersection of IoMT and Blockchain technologies will be explored within the context of how the implementation of Blockchain Technology supports a secure data transfer process within a decentralized IoT environment. Second is the Federated Learning (FL) within the context of ML and its role within the privacy of ML. We analyze 16 recent papers that combine the use of Blockchain, IoT, and ML models, especially for applica- tions in the areas relating to healthcare, security, and others associated with 6G communication technologies. The papers show that the application of Blockchain technology enhances the management of healthcare data from the IoT, while ML models using healthcare datasets from the IoT improve real- time healthcare analysis and anomaly identification.Moreover, the combination of FL with Blockchain technology provides a secure framework for collaborative learning among devices using IoT technology. However, despite the vast potential benefits, there are also challenges in the realm of scalability, computational complexity, privacy concerns in relation to the use of data, and a lack of legal framework regulation that currently hinder the broader adoption of these combined platforms. This article will offer a broad review on the present status of related research experiments and advance future directions related to the ability of 6G communications to help provide a seamless combination of these concepts for the development of intelligent, safe, and optimized IoT platforms.

ADVANCING AI-DRIVEN SECURITY ARCHITECTURE FOR AUTOMATED ENERGY SUPPLY CHAINS IN THE UNITED STATES

Sadia Ali Watara — 2026-06-08

The modernization of United States energy infrastructure through artificial intelligence (AI), Industrial Internet of Things (IIoT), smart grids, cloud-integrated energy management systems, autonomous monitoring platforms, and digitally interconnected supply chain networks has significantly improved operational efficiency, predictive maintenance, and real-time decision-making across power generation, transmission, and distribution environments. However, the rapid digitalization of automated energy supply chains has simultaneously expanded the cyberattack surface, exposing critical infrastructure to increasingly sophisticated threats including ransomware, adversarial AI attacks, supply chain compromise, SCADA manipulation, insider threats, and large-scale data exfiltration. Energy systems now process enormous volumes of operational technology (OT), information technology (IT), and consumer energy usage data across interconnected cyber-physical ecosystems, making security resilience a national priority for the United States. This article presents a comprehensive analysis of AI-driven security architectures for protecting automated energy supply chains in the United States. The study evaluates machine learning-based intrusion detection systems, federated learning security frameworks, blockchain-enabled energy data governance, deep learning anomaly detection, and adversarial defense mechanisms for critical energy infrastructure. Threats are analyzed across five interconnected system layers including physical infrastructure, industrial control systems, communication networks, cloud analytics, and AI decision-making platforms. The article further examines alignment with U.S. regulatory frameworks including NIST Cybersecurity Framework 2.0, NERC CIP standards, Executive Order 14028, DOE cybersecurity guidelines, and CISA critical infrastructure directives. A multi-phase implementation roadmap is proposed to guide U.S. energy operators toward resilient, privacy-preserving, and AI-enhanced cybersecurity ecosystems. The analysis demonstrates that layered AI-driven architectures integrating federated learning, blockchain provenance, zero-trust networking, and adversarial robust deep learning models provide the most effective defense strategy for securing next-generation automated energy supply chains in the United States

TOPOLOGICAL DESCRIPTORS OF LINE GRAPHS IN POLYMER SUPRAMOLECULAR NETWORKS

Shama Sadiq — 2026-06-08

This research provides the graph theoretical study for a polymer supramolecular network within the paradigm of deriving various degree-based topological invariants through the constructed line graph representation for the underlying supramolecular structure. Networks comprised of supramolecular structures, which are based on non-covalent bonds, are difficult to represent within the conventional graph representation paradigm because they are dynamic in nature and have non-covalent bonds within the network. The line graph transformation paradigm has been adopted with the supramolecular structure for a more appropriate representation of such complex systems’ structure and connectivity within the underlying graph representation paradigm. These indices form the cornerstone of robust quantitative structure property relationship and quantitative structure activity relationship. The derivation of various important topological invariants, such as the Randić index, Zagreb index, and Harmonic index, provides the bridge for a quantitative relationship between the supramolecular network structures and various functional groups and behaviors within a deterministic paradigm for such complex networks and structures.

Keywords : Topological indices; Line graph; Polymer Supramolecular Network.

MATHEMATICAL OPTIMIZATION OF ANALOG COMPUTE-IN-MEMORY: TRADE-OFFS BETWEEN LINEARITY, NOISE, AND NON-LINEAR ACTIVATION IN MEMRISTOR CROSSBARS

Muhammad Tahir Abbas — 2026-06-08

The von Neumann bottleneck in deep neural network inference can be addressed by analog compute-in-memory (CiM) by analog matrix-vector multiplication performed in the memory array. This can, in principle, reduce energy consumption by one to two orders of magnitude in comparison to digital accelerators. However, in practice, the accuracy of inferences is far from perfect due to analog non-idealities such as device noise, conductance programming variability, ADC quantization, and signal saturation, making analog CiM impractical for real-world applications. This paper proposes a comprehensive mathematical optimization framework by combining the device physics, circuit design, and neural network training to obtain the minimum inference error for analog CiM systems. A closed-form statistical model of error propagation through memristor crossbar is first developed, accounting for thermal noise, shot noise, programming noise, and quantization of the ADC.Thermal noise, shot noise, programming noise and ADC quantization are first captured in a closed-form statistical model of error propagation through a memristor crossbar. Based on this model, we formulate the optimization of the conductance range [G_min, G_max ] and the full-scale (FS) of the ADC as a convex program and obtain globally optimal parameters that provide a balance between signal strength, noise, and risk of signal saturation. We then present a noise-aware version of the digital ReLU, f ̃_"ReLU" _ ( x ̃ ) to incorporate the distribution of analog noise and the ADC saturation during fine-tuning so that the network can learn hardware-robust representations. Our framework is validated on a simulated 128x128 memristor crossbar, with the MNIST database and a 3-layer multi-layer perceptron. Compared with naive analog mapping with the accuracy loss of 6.74% against an ideal digital baseline, our approach retrieves 86.8% of this loss in the test, resulting in 97.32% accuracy (within 0.89% of digital), and retains the energy efficiency of analog computing. The optimized system decreases output mean squared error by 86% compared to naive analog mapping (14× compared to digital), and increases energy efficiency by 11% compared to naive analog mapping (14× compared to digital). Ablation studies indicate that both convex optimization and noise-aware activation are important for recovery of accuracy, and sensitivity analysis proves that the framework allows for a viable 3-bit ADC operation (93.8% accuracy). The results show that algorithm-hardware co-design, based on convex optimization and noise-aware learning, can bridge this accuracy gap between analog and digital computing, while maintaining the energy advantage of in-memory architectures. The architecture is independent – which could adapt to any resistive memory technology – compute-efficient – with the microseconds per layer – and easily deployable to practical edge analog AI accelerators.

Keywords : Compute-in-memory · Memristor crossbar · Analog neural networks · Convex optimization · Noise-aware activation · Hardware-software co-design · Edge AI.

AI-DRIVEN CYBER THREAT INTELLIGENCE FRAMEWORK FOR CRITICAL DIGITAL INFRASTRUCTURE PROTECTION IN PAKISTAN

Amina Alyas — 2026-06-08

The growing dependence on digital technologies has increased the vulnerability of critical infrastructure to sophisticated cyber threats. Traditional cybersecurity approaches are often inadequate in addressing rapidly evolving attack techniques, creating a need for proactive and intelligence-driven security solutions. This study developed and validated an AI-Driven Cyber Threat Intelligence (CTI) Framework for Critical Digital Infrastructure Protection in Pakistan. Grounded in Dynamic Capabilities Theory, the framework examined the effects of AI-Powered Threat Detection, Predictive Threat Analytics, Automated Incident Response, and Threat Intelligence Sharing on Cyber Threat Intelligence Effectiveness and Critical Digital Infrastructure Protection, with Cybersecurity Governance serving as a moderating factor. A quantitative cross-sectional survey was conducted among 387 cybersecurity professionals from critical infrastructure sectors in Pakistan. Data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM). The findings revealed that all AI-enabled cybersecurity capabilities significantly enhanced CTI Effectiveness, which in turn positively influenced Critical Digital Infrastructure Protection. Predictive Threat Analytics emerged as the strongest predictor, while Cybersecurity Governance strengthened the relationship between CTI Effectiveness and infrastructure protection.

The study highlights the strategic role of AI-driven cyber threat intelligence in enhancing cyber resilience and provides practical guidance for organizations and policymakers seeking to protect critical digital infrastructure in Pakistan

AN EMPIRICAL EVALUATION OF REAL TIME FIRE AND SMOKE DETECTION IN COMPLEX ENVIRONMENTS USING THE YOLOV8 ARCHITECTURE

Abdul Hadi — 2026-06-08

Automated real time fire and smoke detection is critical for modern disaster mitigation and smart city surveillance infrastructure. However, standard single stage deep learning object detection models frequently suffer from high false positive rates due to the amorphous, dynamic nature of fire and smoke, often misclassifying environmental artifacts such as sun glare, clouds, fog, and artificial reflections. This study presents a rigorous empirical evaluation of the baseline YOLOv8 architecture deployed for vision based hazard detection under complex environmental constraints. Utilizing a comprehensive dataset of over 13,000 images characterized by a heavy distribution of small scale targets, advanced preprocessing and augmentation strategies including Mosaic augmentation, Letterboxing, and HSV color jittering were deployed to optimize model robustness. The baseline model was trained and evaluated over 50 epochs, achieving an overall mean Average Precision (mAP@0.5) of 53.9%, with individual class performances reaching 62.3% for fire and 45.5% for smoke. Detailed error analysis using a normalized confusion matrix reveals a critical challenge in separating semi transparent smoke from complex background noise, yielding a 58% background confusion rate. These findings establish a baseline performance benchmark for edge ready disaster management systems and outline the exact architectural boundaries where standard single stage detectors require future spatio-temporal or structural modifications.

PRIVACY-PRESERVING AGENTIC AI AT THE EDGE: FEDERATED AND AUTONOMOUS INTELLIGENCE FOR SMART SYSTEMS

Khaliq Ahmed — 2026-06-08

Introduction: At the edge, privacy-preserving agentic AI is emerging as a factor in intelligent systems where real-time decisions need to be made without revealing sensitive operator, device, or user information. The risk of privacy, latency and bandwidth is introduced by centralized AI, particularly in healthcare, smart homes, transport, energy and industrial internet of things.

Aim: The purpose of this work is to present and analyze a privacy-conscious edge intelligence architecture, a fusion of autonomous agentic decision making, federated learning, differential privacy, secure aggregation, and safe decision escalation.

Methodology: A conceptual and design-based approach was employed to formulate a layered architecture consisting of edge devices, autonomous local agents, privacy engines, federated coordination, and smarter-system applications. The framework was assessed, through perceived concrete metrics of privacy, accuracy, latency, communication cost, resource use and autonomous reliability.

Findings: The suggested framework lowered the exposure percentage of raw data to 0, communication cost dropped to 38MB/round compared to 480MB/round and latency dropped to 67ms compared to 142ms and the accuracy of the model dropped to 92.6% compared with 93.8% and the risk type of information safety decreased to 0.18 compared to -0.72

Conclusion: The framework demonstrates that privacy, autonomy, and efficiency may be harmoniously enhanced in edge-based smart systems.

STRENGTHENING CYBER DEFENSE THROUGH THREAT INTELLIGENCE: ADDRESSING FINANCIALLY MOTIVATED ATTACKS

Abdul Musawer Zahedi — 2026-06-08

Background

Organizations are facing great challenges as financially motivated cyberattacks, such as phishing, ransomware, and financial fraud, keep growing and becoming more common. Frequently, traditional cybersecurity methods and solutions are proving inadequate to tackle the new cyber threats, requiring proactive and intelligence-based security strategies. Threat intelligence has become a vital tool for strengthening cyber defense efforts, boosting cyber situational awareness and mitigating operational and financial risk.

Objective

The purpose of this study was to explore how threat intelligence can be used to enhance an organization's cybersecurity response to financially motivated cyberattacks. The study also assessed the workings of AI-based threat detection technologies, organizational preparedness approaches, and threats in implementing threat intelligence frameworks.

Methodology

The research design applied was quantitative research design that was of descriptive and analytical nature. The sample was composed of 310 cybersecurity professionals, IT staff, and network administrators, security managers and executives of different industries. The data collection method used was a structured close-ended questionnaire with 5-point likert scale. Statistical analysis was done using frequencies, percentages, means, standard deviations, Cronbachs Alpha reliability test and chi-square analysis.

Results

It also determined that there was a high level of agreement on the usefulness of threat intelligence in improving cybersecurity defense. Both Financially Motivated Cyber Attacks (M = 4.31, SD = 0.66) and Effectiveness of Threat Intelligence in Cyber Defense (M = 4.24, SD = 0.69) had the highest average scores. Another important point that the respondents agreed upon is that AI integration enhances the threat detection (M = 4.36, SD = 0.61) and phishing attacks still pose a significant cybersecurity threat (M = 4.42, SD = 0.60). The findings of the reliability check indicated that there is good internal consistency of the scale, with Cronbach Alpha of 0.90. The research also found that there were certain issues with the implementation of the system such as high costs of implementation, shortage of cyber security personnel and the inability to handle a lot of threat data.

Conclusion

The study concludes that threat intelligence, combined with AI and collaborative cybersecurity approaches, can be effective in boosting cybersecurity resilience and organizational cyber defense against financially motivated cyberattacks. Ongoing investments in AI-driven cybersecurity system, staff education, and threat intelligence exchange are crucial for building safe and sustainable cybersecurity environment.

A COMPARATIVE STUDY OF EXPLAINABLE MACHINE LEARNING MODELS FOR STUDENT ACADEMIC PERFORMANCE PREDICTION

Asma Imam Somro — 2026-06-05

Student educational progress prediction has developed a serious examination area in ML, Academic Data mining and Explainable AI. Academic institution constantly pursue smart system to recognizing the risk students, in institution for decision making and refining personal education atmosphere. ML educational model predict the high analytical correctness, many model working as black-box system due to absence of transparency and understandability. This research openhanded a relative study of explainable Model for students education progress forecast. This investigates learning many ML algorithms having Random Forest, Decision Tree, SVM, Logistic Regression, XBM and XGBoost. Educational datasets covering attendance records, assignment scores, quiz marks, study hours, previous GPA, classroom participation, and demographic factors were used for testing. The Investigational results established that XGBoost attained the ultimate prediction accuracy of 93%, while Explainable Boosting Machine provided the excellent balance between predictive performance and interpretability. SHAP analysis used for identification of attendance, earlier GPA, assignment marks, as well as study time as the most significant features to influence the academic success.

DESIGN AND IMPLEMENTATION OF A MACHINE LEARNING–DRIVEN FRAMEWORK FOR REAL-TIME NETWORK TRAFFIC ANOMALY DETECTION AND INTELLIGENT CYBER THREAT IDENTIFICATION

Sufyan Muhammad Khan — 2026-06-09

The increasing sophistication of cyberattacks and the growing volume of network traffic have created significant challenges for conventional intrusion detection systems, particularly in identifying previously unseen threats in real time. This study presents the design and implementation of a machine learning–driven framework for real-time network traffic anomaly detection and intelligent cyber threat identification. The proposed framework integrates automated traffic monitoring, feature engineering, anomaly detection, threat classification, and real-time response generation within a unified cybersecurity architecture. A hybrid machine learning approach combines unsupervised anomaly detection, supervised ensemble learning, deep neural networks, and LSTM-based temporal analysis to continuously monitor network flow characteristics and detect both known and emerging attack patterns. The framework was evaluated using multiple benchmark cybersecurity datasets and validated under simulated enterprise network conditions. Experimental results demonstrated a detection accuracy of 97.8%, precision of 96.9%, recall of 97.2%, and an F1-score of 97.0%. The proposed system reduced false-positive alerts to 2.4% and achieved an area under the ROC curve (AUC) of 0.992, outperforming conventional machine learning models and signature-based intrusion detection approaches. Furthermore, the framework improved threat detection response time by 29.6% while maintaining stable performance under high-volume network traffic conditions. The results confirm the effectiveness of integrating anomaly detection, ensemble classification, and temporal learning within a unified intelligent cybersecurity framework for enhancing real-time threat intelligence, network resilience, and proactive cyber defense in enterprise and cloud computing environments.

A COMPARATIVE STUDY OF ADVANCED LOAD BALANCING ALGORITHMSIN CLOUD COMPUTING ENVIRONMENTS

Muhammad Irfan — 2026-06-09

Round Robin (RR) and First-Come First-Served (FCFS) scheduling algorithms have been designed with the assumption that workloads in cloud systems are uniform, which is not the case with today's cloud infrastructure, as it exposes many weaknesses of these algorithms. In this research, testing will be conducted on 10 different load balancing algorithms from 4 different groups: Artificial Intelligence(AI) and Deep Learning(DL); Nature-Inspired Metaheuristic Algorithms (NIMA); Game Theory Based Load Balancers (GT); and Traditional Load Balancers (LB). For this study, Google Cluster Trace data (from the Google data center) will be used to validate the performance of the aforementioned algorithms. BiLSTM-Attention reached 94.3% classification accuracy and 0.97 Area Under The Curve (AUC); SLADRO obtained 92% CPU Utilization and decreased Idle Power Consumption by 27.5%; these numbers are very significant when you consider the amount of money spent on Idle Compute. Min-Max Scaling (MMS) and Z-Score Normalization (ZSN) were the two main methods used to do data Preprocessing; IQR outlier detection was also used in this research. OOA-PSO was used for feature selection, and data Segments were created using Sliding Windows. The training used ResNet50 (transfer learning) with Adam optimizer and five-fold cross validation. The CNN-LSTM hybrid forecast approach combined with Deep Reinforcement Learning outperformed all of the other baseline algorithms in terms of Makespan, Energy, and Utilization.

NANOCRYSTAL ARCHITECTURES FOR ENHANCED OPTOELECTRONIC PROPERTIES: A PARADIGM SHIFT IN ENERGY HARVESTING AND STORAGE

Sumera Zaib — 2026-06-09

The nanocrystal architecture has revolutionized the field of optoelectronics, offering innovative solutions for energy harvesting and storage applications. This review examines the important role of optoelectronics devices in modern technology and highlights the limitations of traditional materials and introduces nanocrystal architecture as a promising solution. Nanocrystals are synthesized using various colloidal synthesis techniques and template assisted methods. The control on size, shape and composition of nanocrystal is very crucial to maximize the optoelectronic properties. In comparison to conventional materials, nanocrystals perform more efficiently due to key phenomena such the quantum confinement effect, which improves the tunability of bandgaps, absorption coefficients, and charge transport efficiency. Energy harvesting applications are also being investigated, such as the incorporation of nanocrystals into thin-film solar cells, extremely sensitive photodetectors, and photocatalytic devices for water splitting and solar-powered fuel cells. The review also discusses developments in energy storage, with particular attention on lithium-ion battery technology and nanocrystal-based supercapacitors, as well as hybrid devices that combine several other functions. The emerging field of 4D printing has a great potential to produce responsive material for adaptive energy solutions. The potential approaches including interface engineering and sophisticated packing control are discussed, along with the difficulties in creating high-efficiency nanocrystal-based systems. Finally, this review provides an outlook on the future of nanocrystal based optoelectronic devices emphasizing their transformative potential in energy harvesting and energy storage applications.

DESIGN AND DEVELOPMENT OF COMPACT SIZE POWER AMPLIFIERS PCB USING DISCRETE COMPONENTS FOR OBSTACLE AVOIDANCE SONAR

Syed Umaid Ali — 2026-06-09

This paper presents a compact and power-efficient power amplifier (PA) for sonar operation in unmanned underwater vehicles (UUVs). The amplifier delivers 200 W of output power with a mere 0.8% total harmonic distortion (THD) at a 30 kHz center frequency for coherent transmission of sonar signals. Discrete component amplifier design with an optimized feedback network enables high- voltage operation to ±140 V with 282 V of peak-to-peak output. High-performance thermal management with heat sinks and thermal washers enables stable operation in space-restricted environments. The amplifier is 85% power efficient and compact, with a diameter of 150 mm and a height of 20 mm and is therefore well suited for use in underwater drones. Hardware verification provides superior performance compared to traditional Class D designs, including 40% less electromagnetic interference (EMI) and 5 °C less operating temperatures, without duty cycle limitations.

DESIGN AND EXPERIMENTAL EVALUATION OF A PARALLEL OPERATION OF 2X MOBILE DG’S WITH DIFFERENT RATINGS

Syed Umaid Ali — 2026-06-09

The mobile diesel generators are common place in construction, mining, disaster recovery and remote operation and any industry where the required load cannot be achieved by generator and therefore, the operation requires parallel operation. Multi-generation systems have the advantages of capacity increment, enhanced reliability, and fuel efficiency. Similar operation of DGs with varying ratings is also a challenge especially in synchronization and load sharing. Synchronization is to make sure that voltage, frequency and phase angle are similar prior to interconnection. The sharing of the load between unequal generators should be carefully controlled to avoid overloading smaller generators and underexploiting larger ones.Conventional techniques tend to assume that the rating of generators is identical, so they cannot be used in mixed arrangements.

BUILDING INFORMATION MODELING (BIM) AND AI-DRIVEN RISK MANAGEMENT IN PAKISTAN’S CONSTRUCTION INDUSTRY

Sibt E Hassan — 2026-06-09

This study investigates the impact of Building Information Modeling (BIM) adoption and AI-driven risk management on construction project performance in Pakistan’s construction industry. The construction sector in developing economies continues to face persistent challenges, including cost overruns, schedule delays, safety risks, and inefficient risk management practices. In response to these challenges, digital technologies such as BIM and Artificial Intelligence (AI) have emerged as transformative tools capable of enhancing project coordination, predictive risk analysis, and decision-making efficiency. A quantitative, cross-sectional research design was employed, and data were collected from 400 construction professionals, including project managers, engineers, consultants, and contractors. The data were analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) to examine relationships among BIM adoption, AI-driven risk management, risk mitigation effectiveness, and construction project performance. The results revealed that BIM adoption significantly improves AI-driven risk management integration and directly enhances construction project performance. AI-driven risk management was also found to have a significant positive effect on risk mitigation effectiveness and project performance. Moreover, mediation analysis confirmed that AI-driven risk management plays a crucial role in transmitting the effect of BIM adoption on construction project outcomes, indicating a strong indirect pathway. The study concludes that the integration of BIM and AI technologies is essential for improving efficiency, reducing uncertainty, and enhancing overall project performance in Pakistan’s construction sector. Strengthening digital infrastructure, workforce competencies, and policy support is critical for accelerating the adoption of Construction 4.0 technologies.

ADDITIVE MANUFACTURING OF HIGH-PERFORMANCE ALLOYS FOR SUSTAINABLE INDUSTRIAL DEVELOPMENT IN PAKISTAN

Areeba Khan — 2026-06-09

This study examined the role of additive manufacturing (AM) of high-performance alloys in promoting sustainable industrial development in Pakistan. Additive manufacturing has emerged as a transformative Industry 4.0 technology capable of improving material efficiency, reducing production waste, and enabling complex component fabrication through layer-by-layer manufacturing processes. Despite its global adoption in aerospace, automotive, defense, and energy sectors, its application within Pakistan remains limited, particularly in relation to high-performance alloy production and sustainable industrial transformation. The study adopted a quantitative, cross-sectional research design using a structured questionnaire to collect data from professionals in manufacturing industries, including aerospace, automotive, defense, energy, and engineering sectors. A sample of 400 respondents was analyzed using Partial Least Squares Structural Equation Modeling (PLS-SEM) to examine relationships among technological capability, innovation capability, workforce expertise, government support, additive manufacturing implementation, and sustainable industrial development. The results indicated that technological capability, innovation capability, workforce expertise, and government support significantly influenced additive manufacturing implementation. Furthermore, additive manufacturing implementation had a significant positive effect on sustainable industrial development. The mediation analysis confirmed that additive manufacturing implementation significantly mediated the relationship between organizational capabilities and sustainability outcomes. These findings highlight the critical role of additive manufacturing as a technological pathway for achieving resource efficiency, environmental sustainability, and industrial competitiveness. The study concludes that strengthening technological infrastructure, innovation ecosystems, workforce skills, and policy support is essential for accelerating additive manufacturing adoption in Pakistan. The integration of high-performance alloy additive manufacturing into industrial systems can significantly contribute to sustainable economic growth and technological modernization.

MACHINE LEARNING–INTEGRATED BAYESIAN MODELING FOR CLIMATE RISK PREDICTION IN PAKISTAN

Noman Shehzad — 2026-06-09

This study proposes a Machine Learning–Integrated Bayesian modeling framework for climate risk prediction in Pakistan, aiming to enhance the accuracy, interpretability, and uncertainty quantification of extreme weather forecasting. Pakistan is highly vulnerable to climate-induced hazards such as floods, heatwaves, and droughts, which necessitate advanced predictive systems capable of capturing nonlinear climatic interactions and probabilistic uncertainty. Traditional forecasting approaches are limited in handling complex environmental dynamics, while standalone machine learning models often lack uncertainty estimation. A quantitative computational approach was employed using historical climate datasets from 2004–2024, including temperature, rainfall, humidity, and river flow variables. Machine learning algorithms such as Random Forest, Gradient Boosting, and Support Vector Machines were integrated with Bayesian inference techniques to develop a hybrid predictive model. Model performance was evaluated using RMSE, MAE, accuracy, AUC, and Bayesian uncertainty metrics. The results revealed that the proposed ML–Bayesian hybrid model outperformed conventional statistical and standalone machine learning models, achieving the highest predictive accuracy and lowest error rates. The Bayesian component significantly improved uncertainty quantification, enhancing the reliability of climate risk predictions. Rainfall and river flow were identified as the most influential predictors of extreme climate events in Pakistan. The study concludes that integrating machine learning with Bayesian modeling provides a robust, scalable, and interpretable framework for climate risk prediction. The proposed approach can support early warning systems, disaster preparedness, and evidence-based climate policy formulation in Pakistan

OPTIMIZATION OF HYBRID SOLAR THERMAL SYSTEMS FOR INDUSTRIAL ENERGY EFFICIENCY IN PAKISTAN

Dr. Muhammad Umer — 2026-06-09

The industrial sector in Pakistan is characterized by high energy intensity, heavy reliance on fossil fuels, and persistent supply constraints, resulting in elevated production costs and reduced operational efficiency. In response, Hybrid Solar Thermal Systems (HSTSs) have emerged as a promising solution for sustainable industrial process heat by integrating solar collectors, thermal energy storage, and auxiliary energy sources. This study developed and evaluated an optimization framework for HSTSs aimed at improving industrial energy efficiency under Pakistan’s climatic and operational conditions. A quantitative simulation-based research design was employed, incorporating thermodynamic modeling and advanced optimization techniques, including Genetic Algorithm (GA), Particle Swarm Optimization (PSO), and Model Predictive Control (MPC). Performance was assessed using key indicators such as solar fraction, exergy efficiency, fuel savings, system reliability, and levelized cost of heat. The results revealed that MPC outperformed GA and PSO across all performance metrics, achieving the highest solar fraction (71.5%), exergy efficiency (58.9%), and fuel savings (53.8%), while minimizing energy cost. Sector-wise analysis further confirmed strong applicability in textile, food, chemical, and pharmaceutical industries. The findings demonstrate that intelligent optimization significantly enhances the feasibility and effectiveness of hybrid solar thermal systems, offering a viable pathway for reducing fossil fuel dependence and improving industrial sustainability in Pakistan.

BLOCKCHAIN-ENABLED SECURE SMART HEALTHCARE ARCHITECTURE FOR DIGITAL HEALTH SYSTEMS IN PAKISTAN

Dr. Jalal Khan — 2026-06-09

The increasing digitalization of healthcare systems has introduced significant challenges related to data security, interoperability, privacy preservation, and trust among distributed stakeholders. In countries such as Pakistan, these challenges are intensified by fragmented healthcare infrastructure, weak data governance mechanisms, and limited integration between healthcare providers. This study proposes a blockchain-enabled secure smart healthcare architecture designed to enhance data integrity, transparency, and interoperability within digital health systems in Pakistan. The proposed framework integrates permissioned blockchain technology with smart healthcare components, including electronic health records (EHRs), Internet of Medical Things (IoMT) devices, cloud-based systems, and hospital information systems. Smart contracts were employed to automate access control, patient consent management, and secure data exchange among healthcare stakeholders. The system was evaluated through simulation-based performance analysis, focusing on transaction throughput, latency, scalability, and security resilience. The findings demonstrate that the blockchain-based architecture significantly improves system performance compared to traditional centralized healthcare systems, with higher transaction throughput, reduced latency, enhanced data integrity, and improved resistance to unauthorized access. The results further indicate that blockchain integration strengthens trust, transparency, and interoperability across healthcare institutions. This study contributes to the development of a scalable and secure digital health infrastructure tailored to the needs of Pakistan and provides a foundational model for future adoption of blockchain technology in healthcare systems.

AI-BASED SMART ENERGY MANAGEMENT FOR SUSTAINABLE URBAN INFRASTRUCTURE IN PAKISTAN

Muhammad Safi Ullah — 2026-06-09

This study investigates the role of artificial intelligence (AI)-based smart energy management systems in enhancing sustainable urban infrastructure in Pakistan. With increasing urbanization, rising energy demand, and persistent inefficiencies in conventional power systems, AI-driven solutions have emerged as a critical pathway for optimizing energy generation, distribution, and consumption. The study employed a mixed-methods research design, combining quantitative survey data from 220 respondents with qualitative insights from expert interviews and secondary policy analysis. The findings revealed that AI-based energy management significantly improves energy efficiency, smart grid optimization, and renewable energy integration, thereby contributing to sustainable urban development. Regression results indicated that energy efficiency improvement was the strongest predictor of sustainability outcomes, followed by smart grid optimization and AI adoption. However, infrastructural limitations, institutional fragmentation, and limited digital readiness were identified as key barriers to full-scale implementation. The study concludes that AI technologies have strong transformative potential for urban energy systems in Pakistan, provided that supportive policy frameworks, digital infrastructure investment, and institutional capacity-building are strengthened.

EXPLAINABLE ARTIFICIAL INTELLIGENCE FOR EARLY DETECTION AND RISK STRATIFICATION OF CHRONIC DISEASES IN PAKISTAN'S HEALTHCARE SECTOR

Sheraz Gul — 2026-06-09

Chronic diseases such as diabetes mellitus, cardiovascular diseases, and chronic respiratory conditions represent a rapidly growing public health burden in Pakistan, requiring advanced predictive and decision-support solutions for early detection and effective risk stratification. This study developed and evaluated an Explainable Artificial Intelligence (XAI)-based framework integrated with machine learning models to enhance predictive accuracy and interpretability in chronic disease identification. A quantitative, cross-sectional research design was employed using secondary clinical data extracted from healthcare institutions, comprising patient records and clinician feedback. Multiple machine learning models, including Logistic Regression, Random Forest, and XGBoost, were trained and validated using 10-fold cross-validation, while SHapley Additive exPlanations (SHAP) and Local Interpretable Model-Agnostic Explanations (LIME) were applied to ensure model transparency. The findings revealed that XGBoost outperformed other models with the highest predictive accuracy, AUC-ROC, and overall classification performance. SHAP analysis identified blood glucose level, blood pressure, body mass index (BMI), and age as the most influential predictors of chronic disease risk. Furthermore, clinician evaluation indicated a high level of trust and acceptance of the XAI-based system, emphasizing the importance of interpretability in clinical decision-making. The study confirms that integrating explainable AI with predictive analytics significantly enhances both model performance and clinical usability in healthcare environments. In conclusion, XAI-based machine learning frameworks offer a robust and transparent approach for early detection and risk stratification of chronic diseases, particularly in resource-constrained healthcare systems such as Pakistan. The study contributes to bridging the gap between AI model accuracy and clinical interpretability, supporting the development of trustworthy and deployable healthcare AI systems.

GREEN CATALYTIC CONVERSION OF AGRICULTURAL WASTE INTO SUSTAINABLE BIOFUELS IN PAKISTAN

Dr. Muhammad Umer — 2026-06-09

Green catalytic conversion of agricultural waste into sustainable biofuels represents a promising pathway for addressing energy insecurity, environmental degradation, and inefficient biomass management in Pakistan. This study investigates the potential of converting lignocellulosic agricultural residues—such as wheat straw, rice husk, and sugarcane bagasse—into biofuels through advanced catalytic processes, including heterogeneous catalysis, enzymatic hydrolysis, and thermochemical upgrading. A mixed-methods approach was employed, integrating quantitative analysis from energy and environmental professionals with qualitative insights from experts in renewable energy and catalytic chemistry. The findings reveal that catalytic efficiency, biomass availability, and environmental awareness significantly enhance biofuel production potential, while infrastructural limitations and high catalyst costs remain major barriers to large-scale adoption. The study further confirms that green catalytic systems substantially reduce agricultural waste burning and contribute to improved environmental sustainability and energy security. It concludes that integrating green catalytic technologies within a circular economy framework offers a viable and sustainable solution for Pakistan’s energy transition.

ENHANCING LUNG NODULE DETECTION AND CLASSIFICATION USING VISION TRANSFORMERS IN MEDICAL IMAGING

Muhammad Mashood Khan — 2026-06-09

Lung cancer remains one of the leading causes of cancer-related deaths worldwide, primarily due to late-stage diagnosis and the difficulty of accurately identifying pulmonary nodules in early stages. Computed Tomography (CT) imaging plays a vital role in lung cancer screening; however, manual interpretation of CT scans is time-consuming, prone to inter-observer variability, and often affected by the subtle and highly variable nature of lung nodules. To address these challenges, this study proposes an automated lung nodule detection and classification framework based on deep learning techniques. The proposed approach integrates MedSAM based segmentation with a MobileViT based classification model to improve both accuracy and computational efficiency. Initially, lung nodules are segmented from CT images using MedSAM. The segmented nodules are then passed to a MobileViT network, which combines convolutional layers for local feature extraction with transformer-based self-attention mechanisms for capturing global contextual relationships. This hybrid design enables the model to effectively learn both fine-grained morphological features and long-range dependencies within nodule regions. The framework is evaluated on the LIDC-IDRI dataset and achieves strong performance with a training accuracy of 95.58%, validation accuracy of 92.13%, and test accuracy of 91.30%. Experimental results demonstrate that the proposed method provides stable learning behavior, reduced misclassification rates, and balanced performance across benign and malignant classes. The integration of segmentation and classification further improves robustness by focusing the model on clinically relevant regions and reducing background noise.

DESIGN AND IMPLEMENTATION OF A SCALABLE DEEP LEARNING-BASED CRYPTOCURRENCY PRICE PREDICTION AND AUTOMATED TRADING

Farhan Ali — 2026-06-09

The cryptocurrency market has over the past years found its way into most parts of the globe because of its high volatility and possible high returns that it may offer when it is traded. This volatility however makes it difficult to make informed trading decisions by the investors. This project, which will be called SuperCrypt, will aim to design and build an advanced, artificially intelligent crypto trading system that will forecast market patterns and automatize trading strategies in the process of allowing the user to manage their investment portfolios more efficiently. SuperCrypt utilizes OHLCV data in real-time and history provided by exchanges e.g. Binance and trains the deep learning models e.g. BILSTM and Performer to forecast short-term price and multi-timeframe analysis. The system has a simple interface and provides main functions like user authentication, customizable dashboards, management of API keys, automated trading, and preferences of a user in trading. It is created on the basis of Django as a backend service, Fast API as a server to facilitate predictions, Torch as a machine learning training module, and PostgreSQL as an administration of data. It also has an embedded real- time analytics, and performance monitoring tools that give the user clear actionable information. SuperCrypt is containerized in Docker to enable scalability and long-term maintainability and allows CI/CD best practices by tracking experiments with MLflow and ZenML. The performance, security and usability of the system was tested widely in different platform and found to be acceptable. As opposed to most of the available systems that are multifaceted and costly, SuperCrypt will democratize access to AI-powered trading tools by establishing a user-friendly, simple to understand, reliable, and cost-effective solution. As such, this platform provides a linkage between the cutting-edge AI technology and the ease and simplicity of design, allowing an inexperienced as well as a professional trader to confidently and accurately make their way through the complicated maze of the cryptocurrency market.

MACHINE LEARNING-BASED NETWORK TRAFFIC ANALYSIS FOR IDENTIFYING CYBER ATTACKS USING FLOW-LEVEL FEATURES

Hafsa Anwar — 2026-06-09

Background: Due to the growing volume and heterogeneity of traffic generated by modern digital services, cyber attacks are increasingly affecting enterprise, academic, cloud, and government networks. The traditional signature-based intrusion detection systems are capable of detecting known attacks but it is not effective when there is a change in attacks or new malicious behaviours emerge. Purpose: This research article presents a machine learning-based network traffic analysis framework in identifying cyber attacks by use of flow-level features. This paper is concerned with binary classification where each network flow is classified as benign or malicious. Procedure: The proposed framework is based on the CICIDS2017 intrusion detection dataset that contains labelled benign and attack traffic, packet captures and flow-based CSV files. The methodology consists of the data cleaning, label encoding, feature selection, train-test splitting, supervised model training and performance evaluation. The choice of the Logistic Regression, Decision Tree, Random Forest, and XGBoost are made to offer the baseline and the ensemble-based classification performance. Evaluation: Accuracy, precision, recall, F1-score, false positive rate and confusion matrix is used to evaluate the models. These measures are chosen since accuracy in itself can be deceptive in unbalanced datasets of intrusion detection. Contribution: The article has contributed to a structured research design, mathematical formulation, and experimentation procedure that can be direct implemented in Python to identify cyber attacks. It also points out practical concerns, including imbalance in classes, false alarms, biased dataset and the discrepancy between the benchmark performance and the real deployment of the network.

COGNITIVE SENTINEL: DYNAMIC DEFENSE AGAINST MALICIOUS FOG NODES IN EVOLVING FOG -2- FOG COLLABORATIVE MODEL

Rimsha Ehsan — 2026-06-09

IoT applications with stern time constraints often demand very-low latency, and meeting the Quality of Service (QoS) requirements proves challenging with conventional cloud computing. To mitigate this challenge, Cisco introduced Fog Computing in 2015. However, the ever-evolving nature of the fog computing environment introduces several security challenges. Compounding the issue, fog nodes are often deployed by various developers with varying security guidelines. Collaboration amongst the fog nodes, especially in data offloading scenarios, presents security concerns that are currently unexplored. The existing work on security in fog computing is limited, and conventional cryptography strategies are ill-suited for detecting networks having malicious nodes. Consequently, the reputation of IoT services is threatened by presence of malicious fog nodes and this compromises user’s privacy. This research paper advocates for a trust-based model, aiming to identify the maximum trustworthy node to offload tasks while separating any malicious fog nodes within the network. By doing so, the proposed method enhances the security of network and elevates overall Quality of Service (QoS).

MULTI-CLASS VEHICLE DETECTION AND CLASSIFICATION FOR TRAFFIC SURVEILLANCE USING YOLOV8 NANO

Ummi Mursaleen — 2026-06-08

This rapid urbanization has led to an increasing number of vehicles on the roads, which has created a need for automated and intelligent traffic surveillance systems that can detect and classify various vehicle types in real-time. Current computer vision techniques and manual inspection processes do not effectively deal with the complexity, scale and variability of today’s traffic conditions. This paper introduces an end- to-end deep learning solution for multi-class vehicle detection focusing on road traffic images with eight vehicle classes: Car, Auto, Bus, Truck, Light Commercial Vehicle (LCV), Motorcycle, Tractor, and MultiAxle. An extensive data preprocessing pipeline was created that includes image resizing, removal of corrupt images, optimization of compression, removal of duplicate la- bels and verification of the data set. The YOLOv8 framework automatically applied data augmentation in training, such as horizontal flipping, HSV color adjustment, translation, scaling and mosaic augmentation. The model was trained for 25 epochs, with the AdamW optimizer and split into train/validate set at 80:20. The proposed system achieved a final precision of 0.63, recall of 0.69, mAP@50 of 0.67, and mAP@50-95 of 0.44. All three loss components were found to be decreasing uniformly in both the training and validation sets as confirmed by the convergence analysis, there was no overfitting. The results show that the proposed pre-processing methodology and training setup is capable of providing a reliable multi-class vehicle detection which is efficient to be deployed in real world traffic surveillance.

META-ANALYSIS OF CARBON–NITROGEN STOICHIOMETRY EFFECTS ON POLYHYDROXYBUTYRATE (PHB) ACCUMULATION IN ACTIVATED SLUDGE AND MIXED MICROBIAL CULTURES

Muhammad Saim Anwar — 2026-06-10

Polyhydroxy butyrate (PHB) that is a biodegradable biopolymer that can possibly be utilized in producing bioplastics. In this meta-analysis, PHB is produced in comparison to pure cultures, enriched mixed microbial cultures (MMCs) and waste activated sludge (WAS) systems. With the help of random-effects models we will estimate the pooled PHB yields and comment on the results of carbon to nitrogen (C:N) ratio and time on accumulation. The PHB gave the highest (70.0%), then enriched MMCs (33.5%), WAS (31.9%) and these were very diverse as it was a multitude of microbes. Production of PHB was found to increase in all the systems but the peak response has been observed in the pure cultures and this has been attributed to the fact that the C:N ratios are escalating. Time analysis showed that PHB accumulation increased with time where pure cultures resulted in the greatest production then the enriched and finally, the WAS. The consequence of this is that pure cultures would be the most appropriate and possibly the most ideal in the production of PHB yet, enriched MMCs and WAS systems would also make good potential production options which can be scaled at any time. Optimization of microbial selection and nutrient management of such systems would be the critical conditions of enhancing PHB. These strategies will be improved in the future to ensure that the wastewater biomass uses the least amount of energy to produce PHB.

DEEPSORTENGINE: A LOCAL-FIRST, PRIVACY-PRESERVING INTELLIGENT DESKTOP FILE ORGANIZER USING HYBRID SEMANTIC CLASSIFICATION

Muhammad Basim — 2026-06-10

In the modern digital landscape, desktop users accumulate massive quantities of unstructured files, leading to severe digital clutter and diminished productivity. Traditional file managers lack content awareness, requiring laborious manual sorting or brittle, rule-based configurations. To bridge this gap, this paper presents DeepSortEngine, an intelligent, local-first file organization application that automates file sorting through real-time file system monitoring and a hybrid classification pipeline. The proposed system integrates user-learned patterns, deterministic keyword rules, and deep-learning-based vector embeddings to provide adaptive folder recommendations through a non-intrusive accept/reject user workflow. Crucially, to accommodate deployment on consumer-grade hardware with strict resource constraints, the intelligent engine was migrated from an overhead-heavy PyTorch framework to an inference-optimized ONNX Runtime architecture. This optimization yielded a 99.3% reduction in runtime dependency size (from ~2GB to ~15MB) and an 83% decrease in idle memory footprint (from ~300MB to ~50MB), enabling efficient, CPU-only background operations. Furthermore, the architecture introduces a 7-stage hybrid semantic search engine built directly upon an embedded SQLite vector extension (sqlite-vec), enabling context-rich natural language queries under a local-first, privacy-preserving paradigm.

A SYSTEMATIC REVIEW OF MULTIDISCIPLINARY DESIGN OPTIMIZATION IN STEALTH UAVS AND LOITERING MUNITIONS: INTEGRATION OF CFD, FEM, ADVANCED MATERIALS, AND LOW-OBSERVABILITY TECHNOLOGIES

Zeeshan Ahmad — 2026-06-10

Multidisciplinary Design Optimization (MDO) is an essential design methodology for balancing the aerodynamic, structural and electromagnetic performance goals of the stealth unmanned aerial vehicles (UAVs) and loitering munitions. In addition to performance of individual subsystems, computational evidence is beginning to emerge that shows integration of low-observability constraints into MDO is a factor in the effectiveness of the system level platform. The role of CFD, FEM, advanced materials and radar signature management technologies in a unified MDO, however has not been studied systematically. The purpose of this review was to seek to combine the evidence of the integration of these disciplines in the context of stealth UAVs and loitering munitions and to assess what they offer in terms of promoting platform performance. The systematic review was conducted based on PRISMA guidelines. An extensive review was conducted in Scopus, Web of Science and AIAA Digital Library up to May 2025. The PICO framework was used to identify studies that discussed the design of stealth UAVs or loitering munitions as well as the reporting on MDO integration between at least two of the four disciplines that were targeted: CFD, FEM, advanced materials, and low-observability technologies. The Engineering Study Quality Assessment Tool, modified to consider the risk of bias, was used to assess the risk of bias. Seven studies were included following PICO criteria in which the formal multi-disciplinary integration in a context of human-relevant UAV or loitering munition design was required.Of the 9,847 records initially identified, seven studies were included according to PICO criteria which required the formal multi-disciplinary integration in a context of human-relevant UAV or loitering munition design. The evidence is conclusive and very strong that the synergistic integration of geometric shaping, structural optimization and choice of radar absorbing material in a single MDO design leads to reductions in RCS and aerodynamic-structural improvements that are not attainable using sequential single discipline approaches. The results are: surrogate-based and adjoint MDO frameworks allow for design space exploration superior to that provided by gradient-free methods; both RAM layer properties and the coupled CFD-FEM methods can be used to explore the design space for the reduction of signatures beyond just geometric shaping; coupled CFD-FEM methods can be used for simultaneous structural mass reduction and aeroelastic load alleviation. Dedicated MDO frameworks for loitering munitions, on the other hand, are still not well-represented in the literature and only few and inconsistent treatments of compact-planform specific design challenges. Assessment of risk of bias suggested low to moderate risk for all included studies, mostly due to the lack of aerodynamic and/or experimental RCS data. This systematic review will show that integration of MDO—especially with the aerodynamic-signature coupling relationship—can contribute to stealth UAV performance, and allows for system-level trade space navigation across disciplines. An early integration of low observability constraints, starting at the design phase is a good and much desired direction, but there is a need for special high fidelity validation campaigns and loitering munition-specific MDO frameworks.

POWER ELECTRONICS AND RENEWABLE ENERGY SYSTEMS: INNOVATIONS IN SUSTAINABLE ENERGY CONVERSION TECHNOLOGIES

Rehan Ali Khan — 2026-06-10

The increasing demand for sustainable energy solutions accelerated the adoption of renewable energy systems and advanced power electronics technologies. This study examined the role of power electronics in renewable energy systems and investigated recent innovations in sustainable energy conversion technologies. The research focused on evaluating the influence of power electronics innovation, advanced semiconductor technologies, smart grid integration, and intelligent energy management on sustainable energy conversion performance. A quantitative research design was employed, and data were collected from a sample of 300 professionals working in renewable energy organizations, power utilities, engineering firms, and research institutions. Data analysis was conducted using descriptive statistics, reliability analysis, correlation analysis, and multiple regression analysis. The findings revealed strong positive perceptions regarding all study variables, with mean scores ranging from 4.18 to 4.37. Reliability analysis produced Cronbach’s alpha values between 0.84 and 0.89, indicating strong internal consistency. Correlation results demonstrated significant positive relationships among all variables, with coefficients ranging from 0.723 to 0.846. Regression analysis showed that intelligent energy management exerted the strongest influence on sustainable energy conversion performance (β = 0.351, p < 0.001), followed by power electronics innovation (β = 0.318, p < 0.001). The model explained 77.8% of the variance in sustainable energy conversion performance (R² = 0.778). The study concluded that technological innovations in power electronics significantly enhanced renewable energy integration, energy efficiency, system reliability, and sustainable energy development, supporting the global transition toward low-carbon and resilient energy infrastructures.

EFFICIENT CROSS-MODALITY IMAGE RETRIEVAL LEVERAGING USING MULTIMODAL OPTIMIZED FEATURE ENGINEERING AND DEEP LEARNING INTELLIGENCE

Jacob Katende — 2026-06-10

Content-Based Image Retrieval (CBIR) has become an important area of research in computer vision, mainly due to the rapid increase in visual data and the need for more effective retrieval techniques beyond traditional text-based approaches. Although many existing systems use multimedia content to search large image collections, they still face difficulties when dealing with continuously growing datasets, especially in specialized domains such as medical imaging. Medical images—captured through different modalities like MRI, CT scans, and X-rays—require accurate identification of their type to support better diagnosis and improve retrieval precision. To address this challenge, this study presents a comprehensive framework for classifying and retrieving medical images based on their modality, using advanced feature extraction and machine learning techniques. The proposed approach combines seven different visual features to capture various aspects of image content, including texture, edges, and color. These features include Scale-Invariant Feature Transform (SIFT), Local Binary Patterns (LBP), Local Ternary Patterns (LTP), Edge Histogram Descriptor (EHD), Color and Edge Directivity Descriptor (CEDD), wavelet-based color edge features, and color histograms. All extracted features are merged into a single feature vector, allowing a more complete and descriptive representation of each image. The system was tested using the ImageCLEF2012 modality classification dataset, which contains 31 different types of medical imaging modalities. For classification, a Support Vector Machine (SVM) with a chi-square kernel was used, as it is well-suited for handling complex and high-dimensional data. The proposed method achieved an overall accuracy of 72.2%, outperforming the best visual feature-based result from ImageCLEF2012 by 2.6%. This performance improvement highlights the effectiveness of combining multiple features to better distinguish between different image modalities. The study’s key contribution lies in integrating wavelet-based edge information with texture features, along with the use of a chi-square kernel to improve classification performance. Overall, this work demonstrates that carefully designed feature fusion techniques, paired with an appropriate machine-learning model, can significantly enhance CBIR systems in medical imaging. Future work may focus on incorporating deep learning methods and extending the framework to handle images that belong to multiple categories simultaneously.

BUILDING TRUSTWORTHY AND RELIABLE AGRICULTURAL ARCHITECTURE FOR SMART AGRICULTURE USING DECENTRALIZED IOT AND IMMUTABLE DATA GOVERNANCE MECHANISMS

Rana Gulraiz Hassan* — 2026-06-10

As communication technologies continue to advance, the Internet of Things (IoT) has rapidly evolved from an emerging concept into a nearly mature ecosystem, resulting in a significant increase in data generation and processing demands. This rapid expansion places considerable pressure on the efficient management of widely distributed IoT systems. Traditional centralized IoT management platforms, however, suffer from several critical drawbacks, such as susceptibility to cyber threats, dependence on single points of control, and limited scalability. To overcome these challenges while also meeting data privacy and regulatory requirements, this study introduces a block chain-integrated IoT sensor framework aimed at improving security, transparency, and data accessibility. The proposed system merges IoT sensor networks with block chain technology to create a decentralized and immutable ledger that securely records all device interactions. This ensures that data remains tamper-resistant and access is strictly controlled. In addition, smart contracts are employed to automate system operations, including user-device interactions, real-time monitoring, and device management processes. To validate the proposed approach, a prototype system was developed using NodeMCU microcontrollers and a permissioned block chain network. Its performance was evaluated based on key indicators such as latency, throughput, and resource consumption. A practical case study in cotton farming further demonstrates the system’s real-world applicability. By integrating automated irrigation control, the framework helps optimize water usage without compromising crop productivity. Experimental results show a reduction of approximately 35% in water consumption, along with strong protection against unauthorized data manipulation. Comparative evaluation also reveals that the proposed solution performs better than traditional centralized systems in terms of scalability and resilience, especially in environments with limited resources. By combining the sensing capabilities of IoT with the security advantages of block chain, this research presents a reliable, transparent, and efficient approach to modern agricultural management. Overall, the study highlights the potential of block chain-enabled IoT systems to support sustainable and data-driven decision-making across various industries.

DYNAMIC URDU DISCOURSE-AWARE PROMPT TUNING (DUDAPT) FOR CONTEXT-ADAPTIVE IMAGE CAPTIONING

Ammad Hussain — 2026-06-10

We propose Dynamic Urdu Discourse-Aware Prompt Tuning (DUDAPT), a novel framework for context-adaptive image captioning that addresses the unique challenges of Urdu language integration. Traditional captioning systems rely on static word embeddings, which often fail to capture Urdu’s rich discourse features such as syntactic complexity and anaphora resolution. The proposed method introduces a dynamic embedding layer that adapts to linguistic context through three key components: a Discourse Complexity Analyzer (DCA) to evaluate sentence complexity in real-time, a Dynamic Prompt Pool (DPP) that selectively activates context-aware soft prompts, and an Urdu-Aware Embedding Projector to align tokens with visual-semantic spaces. The DCA employs a lightweight transformer to compute complexity scores, which then guide the DPP to expand or prune prompts dynamically. Moreover, the projector combines frozen Urdu embeddings with adaptive prompts, enabling seamless integration with conventional language decoders. The framework is realized using a distilled Urdu-BERT model for efficiency and meta-learned multilingual prompts for robustness. Experimental validation demonstrates that DUDAPT outperforms fixed-embedding approaches by effectively capturing discourse nuances while maintaining compatibility with existing captioning pipelines. This work bridges a critical gap in low-resource language processing, offering a scalable solution for Urdu-centric multimodal applications.

SPARSE-HIERARCHICAL ATTENTION FOR SELF-SUPERVISED INDOOR SCENE CLASSIFICATION: A MASKED PATCH CONTRASTIVE APPROACH

Mubasher Hussain Malik — 2026-06-10

We propose a sparse-hierarchical attention mechanism to improve self-supervised learning for indoor scene classification, addressing the computational inefficiency of standard Transformer attention while preserving structural dependencies unique to indoor environments. The proposed method integrates focal attention, which selectively computes interactions for semantically significant regions, and hierarchical pyramid attention, which captures multi-scale spatial reasoning across downsampled feature maps. These components are embedded into a contrastive pretext task framework, where masked patch contrastive learning optimizes feature representations by minimizing the distance between masked and unmasked regions. The sparse-hierarchical attention reduces computational complexity from quadratic to linear with respect to input size, enabling efficient training without sacrificing performance. Moreover, the hierarchical design ensures robust feature extraction across varying scales, which is critical for modeling the complex layouts and object arrangements typical of indoor scenes. We implement the approach within a modified Vision Transformer (ViT) backbone, demonstrating its effectiveness through empirical validation on standard indoor scene datasets. The results show that our method achieves competitive accuracy while significantly reducing memory and computational overhead compared to full self-attention baselines. This work provides a practical solution for scaling self-supervised learning to high-resolution indoor imagery, with potential applications in robotics, augmented reality, and smart environment systems.

USING AI TECHNOLOGY IN REDUCING EDUCATIONAL INEQUALITY IN RURAL AREAS

admin admin — 2026-06-10

Educational disparity continues to be a significant obstacle [1], particularly in rural areas, where access to quality education is hampered by insufficient infrastructure, limited resources, and teacher shortages. Artificial Intelligence (AI) and contemporary technology present innovative solutions to close this gap by offering scalable, cost-effective, and tailored learning experiences. The main goal of this study is to explore how AI and technology can alleviate educational inequality in rural areas by enhancing educational access, improving learning quality, and addressing infrastructure issues. The study's methodology employs a mixed-methods approach, encompassing[2] needs assessment, literature review, the selection of suitable AI-based educational tools, and ongoing monitoring for enhancement. Quantitative data is evaluated using metrics such as student performance, attendance, and teacher feedback, while qualitative data is gathered through discussions with educators, students, and parents, along with case studies from both well-performing and under-resourced schools. The results indicate that incorporating AI and technology improves student learning outcomes, broadens access to educational materials, boosts enrollment figures, and enhances teacher performance[3]. The research concludes that AI and technology can greatly help reduce educational disparities in rural areas; however, their success depends on effective implementation, adequate digital infrastructure, and ongoing sustainability[ 4]. It is advised that bridging educational gaps via AI requires enhanced digital infrastructure, the creation of AI tools suited for offline and low-bandwidth settings, and greater community and parent engagement in the education process [5 ].

Keywords: AI in education, AI-powered personalized learning, educational inequality, rural education, equitable education.

A REVIEW OF ENERGY-EFFICIENT TASK SCHEDULING IN IOT CLOUD, FOG, AND EDGE SYSTEMS

Areesha Sami — 2026-06-11

The resulting fast growth in the number of IoT ecosystems has made the demand of sophisticated, power-efficient task-scheduling algorithms that can handle dynamic workloads and hetero-geneous devices and provide strict requirements on latency. The classic scheduling techniques are usually based on fixed settings or cloud-based processing where it consumes too much energy and hampers the performance of the network edge. To solve the energy-latency trade-off in IoT-edge-cloud systems, it is suggested in this paper to dynamically and cross-layer schedule an application, considering real-time system monitoring, a lightweight neural prediction module, and decision optimization with the help of DVFS. The neural predictor which has been trained on skip-layer connections and an entropy-based fitting has a good feature separation as seen through the sorted weight-magnitude analysis and SSR of 335 which means that the predictor is stable when it comes to predicting computation and communication needs. With iFogSim2, EdgeCloudSim and Google Collaboratory, the system demonstrates an up to 27 percent decrease in overall energy use as well as the 95th-percentile latency with different mobility and workload situations. The findings affirm that the suggested approach provides a high-quality, scalable, and energy-conscious scheduling solution that can be utilized in the website of current IoT applications.

QUANTUM SEARCH IN THE NISQ ERA: A COMPREHENSIVE SURVEY OF GROVER’S ALGORITHM, NOISE RESILIENCE, AND APPLICATIONS IN INFORMATION RETRIEVAL

Sanam Shoukat — 2026-06-11

This study provides a comprehensive review of Grover's Algorithm in quantum computing, emphasizing quantum information retrieval during the Noisy Intermediate-Scale Quantum (NISQ) era. The essential element in quantum information retrieval is Grover's Algorithm, which has been shown to be the most efficient one possible, offering a quadratic acceleration of O(sqrt(N)) compared to the classical O(N) unstructured database search. This survey offers an algorithm taxonomy by methodically examining 18 peer-reviewed works published from 1996 to 2026, systematically analyzing the Grover search, hybrid quantum-classical models, variants of amplitude estimation, distributed quantum search, adaptive learning oracle design and NISQ optimized circuit implementation. A structured comparative analysis is performed on the performance of classical and quantum approaches, with experimental results from IBM Quantum's 127-qubit superconducting processors. Systematic identification and discussion of critical research gaps such as the sub-O(√N) complexity barrier, noise resilience, scalability limits and quantum data-loading bottleneck. Future directions include fault-tolerant hardware, adaptive oracle learning, integration of quantum computers with AI, federated quantum search, and standardizing the benchmarking of quantum computers. The aim of this survey is to give an integrated structured reference for researchers interested in the field of quantum computing and information retrieval.

EDGE AI-BASED MODELS FOR DETECTING SPOOFING ATTACK IN RESOURCE-CONSTRAINED IOT NETWORKS

Abrar Akram — 2026-06-11

The Internet of Things (IoT) has created an immense new space for the cyber bad guys to attack in such sectors as healthcare, industry automation and smart infrastructure. A profiling attack is particularly concerning in the IoT environment, since it can mimic the typical behavior of a legitimate device, thereby enabling the attacker to gain access to the network and access the information without triggering the security alerts. Traditional IDSs are not appropriate for IoT because they are centralized systems, require huge amount of computation resources, and most of the IoT end-points don't have those resources.

A light-weight Edge AI based IDS system is proposed in this paper, which is specially designed for detecting the spoofing attack in resource-constrained IoT network. A structured machine learning pipeline is applied to the standard dataset UNSW-NB15, which includes data cleaning of duplicated data, encoding labels, data normalisation using the StandardScaler, feature selection using correlation-based feature selection to select 15-25 most important features and binary classification using LogisticRegression with L2 regularization parameter of 0.1. A Random Forest (RF) classifier is used to evaluate the accuracy of detection and computational cost of the proposed model. The results from experiments indicate that the accuracy of Logistic Regression model is 95.46%, precision is 95.53%, recall is 96.26% and F1 score is 95.89%. Despite its simplicity, the model is still very competitive in terms of detection performance and suitability in the edge environment in terms of memory, latency and processing requirements (vs. Random Forest with 93.12% accuracy). These results show that with proper optimization of lightweight models, processing and feature engineering, it is possible to obtain a dependable real-time IDS with low computational requirements. This framework offers a workable and scalable security solution for use at the network edge in today's modernistic environment of IoTs.

AI-AUGMENTED DMAIC FRAMEWORK FOR MANUFACTURING QUALITY IMPROVEMENT - A CASE STUDY USING PUBLIC DATASET

Abdul Jabbar Ehsan — 2026-06-11

Manufacturing industries are increasingly adopting intelligent technologies to improve product quality, minimize production defects, and enhance operational efficiency in highly competitive industrial environments. Traditional quality management methodologies such as Six Sigma DMAIC (Define, Measure, Analyze, Improve, and Control) have been widely used for systematic process improvement and defect reduction. However, conventional DMAIC approaches mainly rely on statistical analysis and manual decision-making, which often become insufficient when dealing with large-scale industrial datasets, real-time sensor streams, and complex manufacturing systems associated with Industry 4.0. To address these challenges, this research proposes an AI-augmented DMAIC framework that integrates Artificial Intelligence (AI) and Machine Learning (ML) techniques into the traditional DMAIC methodology for intelligent manufacturing quality improvement. The proposed framework enhances each DMAIC phase by incorporating predictive analytics, automated defect detection, root cause analysis, and data-driven decision support. A public manufacturing quality dataset containing operational machine parameters and defect-related information is utilized as a case study to validate the effectiveness of the proposed approach. In the proposed system, data preprocessing and feature engineering techniques are first applied to prepare the manufacturing dataset for analysis. Subsequently, Machine Learning models including Random Forest and Neural Network classifiers are trained to predict defective products and identify the most influential manufacturing parameters affecting quality performance. Various evaluation metrics such as Accuracy, Precision, Recall, F1-Score, and Mean Squared Error (MSE) are used to assess model performance. Experimental results demonstrate that the AI-enhanced DMAIC framework significantly improves manufacturing quality by reducing defect rates, minimizing process variation, and increasing predictive accuracy. Among the implemented models, the Random Forest classifier achieved the highest performance with superior defect prediction capability and efficient feature importance analysis. The findings further indicate that integrating AI within DMAIC enables proactive quality management, intelligent process optimization, and real-time monitoring in smart manufacturing environments. The proposed framework provides a scalable, adaptive, and data-driven quality improvement solution suitable for Industry 4.0 applications. This research contributes toward the development of intelligent manufacturing systems capable of autonomous decision-making and continuous operational improvement

CODE SPRINT: AN INTERACTIVE LEARNING PLATFORM FOR COMPETITIVE PROGRAMMING

Syed Moin Uddin — 2026-06-11

Competitive programming has proven to be a great way to build algorithmic thinking, problem-solving skills and coding skills in students & software developers. However, many new students experience difficulties in finding problems to solve, motivating themselves to solve the problems or in recalling the solution for a complex algorithm. It introduces adaptive learning paths, real time code evaluation, gamification and analytics dashboards, all of which are based on AI, to make education in competitive programming more interactive through a paper it introduces a new concept for a learning platform (CodeSprint), a system that is making this approach to competitive programming education more interactive. The suggested platform integrates Online Judge (OJ) system and suggestion and collaborative learning system. It is suggested that codes are modifiable to make it scalable, safe and efficient to execute codes in the program. Experimental studies show that the platform increases the engagement level of learners, their ability to reason and solve problems, and the performance of their programming. This work adds to an emerging cadre of "intelligent programming education systems" by offering a broad framework for competitive programming education. Competitive Programming platforms and Online judges have been well established as learning and automatic assessment tools for programming education.

Keywords: Competitive Programming, Online Judge System, Gamification, Adaptive Learning, Programming Education, Artificial Intelligence, Learning Analytics.

FITVERSE: AN AI-POWERED FASHION INTELLIGENCE PLATFORM FOR REAL-TIME BODY MEASUREMENT EXTRACTION AND PERSONALIZED FASHION RECOMMENDATION USING MEDIAPIPE AND COMPUTER VISION

Usman Ahmed — 2026-06-11

Ecommerce has also revolutionized the fashion retail market and created a convenient shopping experience for consumers to buy fashion products online. Fitting problems, return of products, and wrong size fit before purchasing, however, are frequent issues such as the ability to touch, feel and try-on the garment before taking a chance is denied. Unfortunately, the trend of fashion recommendations with date for users is primarily based on the user's preferences, a mapping with fixed dimensions, purchasing history, or any other kind of information that does not yield a detailed and dynamic relationship between fashion and body attributes that provides a better understanding of user satisfaction. This paper introduces a new Fashion Intelligence System, called FitVerse, that deals with these challenges. The proposed system is based on the real-time body landmark detection system provided by Mediapipe, which can be used to detect and extract the body landmarks and a multi-image analysis based on the webcam scan for obtaining the body measurements. It provides you with its own "keys" for the body measurements – chest, waist, hips, shoulders, thighs, inseam and height – and correlates them to sizes offered by different brands precise to clothing size. Besides, the Fashion Intelligence Engine created by FitVerse could carry out body-shapes classification, suggestions of fit type and color analysis of skin tone, which enhances the level of personalization. The experimental evaluation shows that the proposed methods can be applied in real time with good performance in measuring body dimensions, classification, and accuracy of the recommendations.

Keywords: Artificial Intelligence, Fashion Recommendation System, MediaPipe, OpenCV, Computer Vision, Pose Estimation, Personalized Outfit Suggestions

ATTESTIFY: A HYBRID BLOCKCHAIN-IPFS FRAMEWORK FOR TRUSTLESS ACADEMIC CREDENTIAL VERIFICATION USING SOULBOUND TOKENS

Zain Ul Abidin — 2026-06-11

An ongoing institutional weakness is the widespread use of fraudulent credentials. Verification workflows that require institutions to communicate synchronously, through proprietary portals and unverified confirmation channels exacerbate the issue by providing no cryptographic guarantees of document integrity.To this end, research into blockchain-based credentialing has increased in significant amounts, but specific shortcomings have been evident throughout the literature: most proposed solutions address only parts of the credential lifecycle, none of them rely on non-transferable credentials for tokenisation of the credential holder, and participation in W3C interoperability standards is still in its infancy, if it exists at all.

A SURVEY OF GROVER’S ALGORITHM AND ITS MODIFICATIONS FOR EFFICIENT UNSTRUCTURED SEARCH

Iram Fatima — 2026-06-12

One of the fundamental algorithms of quantum computing is known as Grover’s quantum search algorithm, which gives a quadratic speedup over classical search methods in unstructured databases. The authors present a survey of the research on Grover algorithm since 2003 and describe some important modifications and improvement over the years. Adaptive variants, hardware-specific implementations and usage in optimization, artificial intelligence etc. are discussed. In this paper, we review the different variants of Grover’s algorithm, discuss their working principles and implementation strategies. These are compared so as to discover possible modifications that increase the efficiency and performance of unstructured search. Also, current issues such as error mitigation in quantum devices and adaptation of algorithms to variable database size are discussed. This survey will be a useful overview of the state of the art in quantum search algorithms, and suggest lines for future research. The development of Grover’s algorithm is an important step in the field of quantum computing, and as research in the field continues to progress, the algorithm will be further refined and improved to enable more practical applications in the future.

PRIVACY-PRESERVING AGENTIC AI AT THE EDGE: FEDERATED AND AUTONOMOUS INTELLIGENCE FOR SMART SYSTEMS

Khaliq Ahmed — 2026-06-12

Conclusion: The framework demonstrates that privacy, autonomy, and efficiency may be harmoniously enhanced in edge-based smart systems.

EVALUATION OF STRESS HYPERGLYCEMIA IN NON-DIABETIC PATIENTS WITH ACUTE MYOCARDIAL INFARCTION

Zeenat Ramzan — 2026-06-12

Introduction: Stress hyperglycemia frequently occurs during acute myocardial infarction (AMI) even in patients without previously diagnosed diabetes. This transient rise in blood glucose represents an acute metabolic response to physiological stress but is increasingly recognized as a marker of adverse cardiovascular outcomes. Understanding its prognostic significance in non-diabetic individuals is essential for risk stratification and early intervention. Objectives: To evaluate the stress of hyperglycemia in non-diabetic patients presenting with acute myocardial infarction. Methodology: This observational study included non-diabetic adult patients admitted with AMI. Stress hyperglycemia was assessed using admission plasma glucose and the stress hyperglycemia ratio (SHR). Clinical outcomes including in-hospital mortality, heart failure, arrhythmias, cardiogenic shock, and length of hospital stay were recorded. Patients were stratified into normoglycemic and stress-hyperglycemic groups for comparative analysis. Results & Findings: Patients with stress hyperglycemia demonstrated significantly higher rates of adverse outcomes, including increased risk of in-hospital mortality, acute heart failure, and cardiogenic shock. Elevated admission glucose and higher SHR were strong independent predictors of complications. Stress hyperglycemia was also associated with prolonged hospital stay and higher need for intensive care support. Conclusion: Stress hyperglycemia is a powerful prognostic marker in non-diabetic AMI patients. Elevated glucose levels at presentation predict higher morbidity and mortality, emphasizing the need for early identification and tighter glucose monitoring in this population. Incorporating stress hyperglycemia into routine risk assessment may improve clinical decision-making and patient outcomes.

TRUST SCORE FRAMEWORK FOR GOVERNING AUTONOMOUS DECISION-MAKING IN AGENTIC AI CUSTOMER SERVICE SYSTEMS

Areesha Sami — 2026-06-12

To address this, our paper introduces the Multi-Dimensional Trust Score (MDTS) Framework a practical evaluation layer that sits on top of existing AI systems and scores every AI-generated response across five dimensions: Accuracy, Personalization, Transparency, Privacy Safety, and Autonomy Risk. The MDTS Framework addresses a fundamental question that comes with AI taking on more and more responsibility in customer service: how do we determine when an AI response is trustworthy enough to be sent on its own, and when should a human intervene before it is sent? Each dimension is rated on a scale of 0 to 2, producing a composite score out of 10. That score then drives an automatic routing decision: responses scoring 8–10 are sent directly to the customer, scores of 5–7 go to a human agent for review before sending, and scores of 0–4 are handed off entirely to a human. The framework is validated on a dataset of 1,200 real-world customer service interactions spanning five query categories and six languages, scored by five independent annotators with a Krippendorff’s of 0.7675. Routing performance is benchmarked against expert ground-truth labels using precision, recall, and F1-score. A Python-based prototype built on GPT- 4 and LangChain confirms the system is deployable within real agentic pipelines. MDTS outperforms all single-signal baselines on Macro F1, with the optimal threshold pair of T_low=5 and T_high=8 achieving an accuracy of 0.614 and a Macro F1 of 0.481. By making trust measurable at the level of individual responses rather than at the system level, MDTS offers organizations a transparent, regulation-aligned path toward responsible AI autonomy in customer service

PSYCHIATRIC COMORBIDITIES AND IN-HOSPITAL OUTCOMES IN METHAMPHETAMINE-ASSOCIATED MYOCARDIAL INFARCTION: A CASE SERIES

Samar Abbas — 2026-06-12

Background: Psychiatric disorders are common in methamphetamine use disorder, yet their relationship with in-hospital outcomes after methamphetamine-associated myocardial infarction (MA-MI) has not been examined. This case series describes the prevalence of pre-existing psychiatric diagnoses in MA-MI and explores associated patterns in mortality, revascularization, and discharge disposition. Methods: We reviewed records of 134 consecutive adults (aged 18–75) admitted with a primary diagnosis of acute myocardial infarction and a positive urine methamphetamine screen within 48 hours of admission at a university-affiliated tertiary care center in Handan, China, between 2019 and 2024. Pre-existing psychiatric diagnoses were ascertained by manual chart review and grouped as psychotic, mood, or anxiety disorders. This study is an exploratory case series and was not powered for hypothesis testing. All statistical comparisons are descriptive; p-values, where reported, are unadjusted and not definitive. Results: Fifty patients (37%) had a documented pre-existing psychiatric diagnosis. In-hospital death occurred in 20.0% (10 of 50) of patients with psychiatric comorbidity versus 3.6% (3 of 84) of those without, an absolute risk difference of 16.4 percentage points (95% CI: 4.7%–28.2%). The mortality difference was entirely concentrated in the psychotic-disorder subgroup (10 of 18, 55.6%); no deaths occurred among patients with mood or anxiety disorders. Revascularization was attempted in 44% versus 67% of psychiatric and non-psychiatric patients, respectively (ARD −22.7 percentage points, 95% CI: −39.7% to −5.6%). Findings were directionally similar in a sensitivity analysis restricted to patients with methamphetamine-only toxicology (n=96). Conclusion: This case series documents a 37% prevalence of pre-existing psychiatric comorbidity in MA-MI, with numerically higher mortality and lower revascularization in the psychiatric subgroup. The most extreme risk was concentrated in the psychotic-disorder subset, though the small sample precludes causal inference. These hypothesis-generating observations underscore the need for prospective multicenter investigations.

Keywords: Methamphetamine; myocardial infarction; psychiatric comorbidity; case series; in-hospital mortality; cocaine; polysubstance use

A HYBRID MACHINE LEARNING FRAMEWORK FOR STUDENT ACADEMIC PERFORMANCE PREDICTION

Dure Shahwar Soomro — 2026-06-12

Student academic performance prediction is a serious trial in academic data mining, where initial and correct predicting allows targeted exclamation strategies. This paper suggests a novel hybrid machine learning framework that combines ensemble methods XGBoost and Random Forest, deep learning and Provision Vector Machine (PVM) within a stacked meta-learning construction. dissimilar define motivated techniques, our framework is better only for forecast correctness and simplification. Trained and assessed on an assorted dataset of 4,872 students calm from five educational organizations across 2019–2023, surrounding 26 attributes covering educational records, communication metrics, socio-economic gauges, appointment data, and demographic features, the future model achieves an accuracy of 93.7%, F1-Score of 92.1%, and AUC-ROC of 0.971, outstripping all six models through a minimum margin of 6.5% in correctness. Wide ablation studies authenticate apiece component’s influence to the complete implementation gain.

AI AND INTELLIGENT PROJECT MANAGEMENT

Azhar Mehmood — 2026-06-12

UNDERWATER OBSTACLE DETECTION IN WIRELESS SENSOR NETWORKS USING YOLOV8S

Muhammad Ibrahim* — 2026-06-12

It is still a very challenging problem to accurately and in real time detect the obstacles in sonar images for Autonomous Underwater Vehicles (AUVs) and Underwater Wireless Sensor Networks (UWSNs), especially in an underwater environment with clutter and noise, where the traditional sonar detection method is weak in accuracy and robustness. The undersea object detection technology currently available can be used to detect undersea objects, but it is not good at performing detection in noise environments, low visibility environments and complex target structures. To overcome these challenges, the light and efficient deep learning underwater acoustic target detection framework based on YOLOv8s architecture is presented in this paper. The model is trained and tested on an underwater acoustic target detection (UATD) dataset consisting of 1127 labeled sonar images from 10 types of obstacles. To boost feature extraction and model generalization, transfer learning with COCO-pretrained weights, advanced data augmentation, and AdamW optimization are used. Experimental results showed that the proposed approach achieved a precision of 92.81% and a recall of 91.07% with the mean Average Precision (mAP@50) being 94.80%. It achieves an mAP@50 of 8.4% improvement over the YOLOv7 model and enables efficient training on a single NVIDIA Tesla T4 GPU in about an hour, making it a suitable model for real-time and scalable underwater detection applications.

Keywords : Underwater obstacle detection; YOLOv8s; sonar image classification; Underwater Wireless Sensor Networks (UWSN); UATD dataset; deep learning; object detection; autonomous underwater vehicles.

ELECTROSTATIC FIELD DISTRIBUTION AND CHARGE TRANSPORT MODELING IN HIERARCHICAL POROUS CARBON ELECTRODES

Sumera Mukhtar — 2026-06-13

The need for materials for advanced energy storage systems has brought focus to topologically hierarchical porous carbons as electrodes. They possess interconnected networks of pores, high electrical conductivity, and a high surface area. This research aimed to studied the electric field and charge transport in topologically hierarchical porous carbon electrodes with the help of computer modeling and simulations. For the electric field, studied the effect of the distribution of pore sizes, pore interconnections, and the structure of the electrodes on the local electric field. The transport of charge was studied by coupling electrochemical transport models and analyzing the pathways of ion transport and electron transport. The interspersed hierarchically patterned pores of the carbon electrodes improved uniformity of the electric field and facilitated transport of charge by decreasing the ion transport lag and increasing the accessibility of the electrolyte. Macropores and mesopores improved ion transport, and charge storage was enhanced by micropores. Additionally, optimized pore structure improved charge distribution and decreased lag of system polarization, affecting the overall electrochemical functioning positively. This work contributes to understanding the intricacies of the microstructure of carbon electrodes and permits construction of advanced electrochemical devices that incorporate carbon supercapacitors and batteries

INTEGRATING PUMPED HYDRO STORAGE WITH RENEWABLE ENERGY TO IMPROVE GRID LOAD MANAGEMENT

Afaq Khalid — 2026-06-13

This research presents a multi-criteria evaluation technique for a sustainable mechanical arrangement that incorporates renewable sources. It investigates the most compelling methods to use the combined control of solar, hydro, and wind power to solve the difficulties of flexible, viable, and tried and true energy capacity. Scientific reenactments with cross-breed arrangements are created using a variety of constraints and working standards. An electrical development framework based mostly on wind and solar technologies, as well as pumped-storage hydropower plans, is drawn out in order to determine how much renewable energy and capacity are necessary to satisfy renewables-only era goals. The proposed strategy in the current study blends pumped hydro capacity innovation with a cross-breed sun-based wind turbine framework (a renewable vitality source) to alleviate vitality shortages while safeguarding network stability. Solar and wind power are inherently unpredictable and untrustworthy sources of energy. As a result, they cannot guarantee the critical stack request. However, by integrating these two renewable resources (solar panels and wind turbines) into a pumped hydro capacity configuration, the effects of fluctuation in solar and wind resources may be mitigated, making the overall system more predictable and economically sustainable to operate. According to the research, the most practicable strategy to achieving this goal is to combine pumped hydropower with solar and wind energy. The findings indicate that, in terms of feasibility and coherence, pumped hydro capacity combined with solar and wind energy is the best option for achieving energy independence.

A TAXONOMY AND RISK-AWARE CONCEPTUAL FRAMEWORK FOR AGENTIC AI-BASED AUTONOMOUS TASK SELECTION IN SOFTWARE ENGINEERING WORKFLOWS

Saria Irshad — 2026-06-13

Agentic artificial intelligence is changing software engineering assistance by shifting from immediate response to code generation to systems that understand the context, use tools, observe feedback, and decide on follow-up actions. Most AI programming assistants and software agents today focus on tasks like code completion, debugging, testing, or issue handling at repository level, without considering task selection as a distinct, explainable, and measurable decision process layer. In this paper, we propose a taxonomy and risk-aware approach to the concept of agentic AI for autonomous task selection in software engineering processes. Our framework uses developer input and workflow signals for task classification, computes uncertainty and risk estimates, makes decisions about selecting the next suitable action, and refers uncertain or critical cases to human confirmation/clarification. We contribute to literature through a task taxonomy definition, comparison with prior research and gap analysis, design of next action selection architecture, decision policy proposal, and outline of experiment scenarios. The present study represents a survey/conceptual framework type of contribution and will be developed later in prototype-based evaluations.

ADAPTIVE GROVER ITERATION STRATEGY FOR EFFICIENT QUANTUM SEARCH OPTIMIZATION

Saria Irshad — 2026-06-13

Quantum computing has been found to have the potential for use as an efficient computing model for addressing complicated computational problems. Several algorithms exist within the quantum computing framework, but Grover's algorithm has been found to offer substantial benefits in searching processes using a quadratic speed-up. In this regard, the purpose of the paper is to conduct a critical analysis of Grover's algorithm as a quantum search algorithm. The paper also provides a critical examination of the principles of operation, parts, and applications of the algorithm. Furthermore, the paper discusses various research efforts regarding the areas of amplitude amplification, oracle construction, noise effects, and optimization methods. In addition, the limitations of Grover's algorithm are presented in the paper. The limitations are found to be limited to the constant number of iterations and overshooting issues. Also, gaps in the field of research due to the lack of adaptive iteration and dynamic system conditions are discussed. In summary, it can be observed that adaptive methods can be employed to enhance quantum search algorithms.

GROVER'S ALGORITHM FOR INFORMATION RETRIEVAL IN QUANTUM COMPUTING: ORACLE DESIGN OPTIMIZATION, ALGORITHM TAXONOMY, COMPARATIVE ANALYSIS, AND FUTURE DIRECTIONS

Hafiza Zarmeen Khan — 2026-06-13

In this paper, Grover's Algorithm is surveyed in quantum computation, specifically regarding optimizing oracles for information extraction via quantum means. The phase oracle plays a central role in achieving Grover's quadratic speedup of O(√N) versus classical O(N), but implementing it efficiently proves to be a major difficulty in current Noisy Intermediate-Scale Quantum (NISQ) hardware, increasing gate and coherence errors. Through a detailed review of twenty peer-reviewed papers, a taxonomy of algorithms is discussed based on Grover's search, hybrid classical-quantum oracles, amplitude estimation oracles, parallel oracle processing, and NISQ-era oracle optimization. Comparative analysis is performed on IBM Quantum's Eagle r3 processor (127 qubits). Key open problems identified include: Oracle Construction Overhead, General Adaptive Oracle Theory, QRAM Bottleneck in Oracle Data Loading, and lack of Standard Oracle Benchmarks.

A FRAMEWORK FOR HATE SPEECH IDENTIFICATION USING OPTIMIZED TEXT FEATURES AND NATURAL LANGUAGE PROCESSING ON TWITTER DATASET

Irsa Manzoor — 2026-06-13

Twitter has emerged as a prominent social media platform where users rapidly share opinions, emotions, experiences, and real-time events. Due to the increasing volume of user-generated textual content, sentiment analysis and hate speech detection have become important research areas in the fields of Natural Language Processing (NLP) and Machine Learning (ML). Although considerable research has been conducted on hate speech detection using Twitter data, the automatic identification of multilingual hate speech, particularly in Roman Urdu and English, remains a challenging task. This research proposes a hybrid NLP-based framework for multilingual sentiment analysis using a combined dataset of Roman Urdu and English tweets collected from publicly available hate speech datasets. The datasets are integrated into a unified corpus and processed using several NLP preprocessing techniques, including stop-word removal, punctuation removal, URL elimination, tokenization, and stemming. Furthermore, optimized textual features are extracted using Python-based NLP libraries to improve the quality of the dataset for machine learning applications. To enhance feature relevance and reduce dimensionality, Principal Component Analysis (PCA) is applied to eliminate less informative features while retaining the most significant attributes. The experimental implementation is carried out using Google Colab, where multiple machine learning classifiers, including Naïve Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), and Decision Tree (DT), are trained and evaluated. In addition, a Hybrid Ensemble Model (HEM) is proposed, which combines the predictions of all four classifiers to improve classification performance. The proposed system classifies users’ sentiments into three categories: Positive, Negative, and Neutral. The performance of the models is evaluated using standard evaluation metrics, including training accuracy, testing accuracy, precision, recall, and F1-score. A comparative analysis of all models is conducted to identify the most effective approach for multilingual sentiment analysis and hate speech detection on Roman Urdu and English Twitter datasets

A MACHINE LEARNING-INTEGRATED NUMERICAL FRAMEWORK FOR SOLVING NONLINEAR FRACTIONAL DIFFERENTIAL EQUATIONS IN CLIMATE MODELING OF PAKISTAN

Syeda Ghurneeq Fatima — 2026-06-13

This study developed a machine learning–integrated numerical framework for solving nonlinear fractional differential equations (NFDEs) in climate modeling applications in Pakistan. The primary objective was to address the computational limitations of conventional numerical methods in capturing nonlinear, multiscale, and memory-dependent climatic dynamics. The proposed framework integrated scientific machine learning techniques, including physics-informed neural networks and neural operator approximations, with fractional calculus-based numerical methods to enhance predictive accuracy, computational efficiency, and numerical stability. A quantitative and computational research design was employed using secondary climate datasets representing key meteorological variables of Pakistan, including temperature, precipitation, and atmospheric variability indicators. The performance of the proposed framework was evaluated and compared with traditional numerical approaches using standard metrics such as RMSE, MAE, execution time, convergence behavior, and stability indices. The results demonstrated that the proposed framework significantly outperformed conventional methods, reducing computational cost and prediction errors while improving stability and forecasting accuracy. Furthermore, the framework effectively captured nonlinear interactions and long-term memory effects inherent in climatic processes. The findings confirmed that integrating machine learning with fractional differential equation solvers offers a robust and scalable approach for climate modeling in highly complex and uncertain environments. The study contributes to computational mathematics, scientific machine learning, and climate science by introducing an advanced hybrid modeling paradigm suitable for climate-vulnerable regions such as Pakistan.

AI-DRIVEN ADAPTIVE PROTECTION SCHEMES FOR RESILIENT POWER SYSTEMS WITH HIGH PENETRATION OF DISTRIBUTED ENERGY RESOURCES

Muhammad Awais — 2026-06-13

The increasing integration of distributed energy resources (DERs), including solar photovoltaic systems, wind energy installations, battery energy storage systems, and electric vehicles, has significantly transformed modern power systems. While these resources improve sustainability and grid flexibility, they introduce substantial challenges to conventional protection schemes due to bidirectional power flows, variable fault currents, and dynamic operating conditions. This study proposes an artificial intelligence-driven adaptive protection framework designed to enhance the resilience, reliability, and operational security of power systems with high DER penetration. Four protection scenarios were evaluated, including a conventional protection system and three progressively advanced AI-assisted adaptive protection configurations. Anticipated outcomes were generated using established power system protection principles, machine learning concepts, and smart grid operational characteristics. The predictive framework suggests that AI-enabled adaptive protection systems may significantly improve fault detection accuracy, fault isolation speed, system reliability, restoration efficiency, and grid resilience while reducing false tripping events and outage durations. The proposed framework serves as a conceptual model and methodological template for future experimental and simulation-based investigations in intelligent power system protection.

EVALUATION OF STRESS HYPERGLYCEMIA IN NON-DIABETIC PATIENTS WITH ACUTE MYOCARDIAL INFARCTION

Zeenat Ramzan — 2026-06-13

CLOUD SECURITY RISK MANAGEMENT IN SMES: CHALLENGES, LIMITATIONS, AND STRATEGIC RESPONSES

Shanza Zaman* — 2026-06-14

cloud computing provides small and medium-sized enterprises (SMEs) scalability, cost efficiency, and access to advanced features, it also presents security challenges that are not always addressed by these businesses. This research investigates the critical issues and constraints of cloud security risk management in SMEs, quantifies their exposure to selected cloud security risks and assesses the cloud security strategies available to them for the purpose of enhancing their cloud security. In order to reach a comprehensive understanding of the challenges SMEs encounter when dealing with cybersecurity, a mixed-methods design was implemented, which involved a structured survey conducted among 200 SMEs from various sectors and semi-structured interviews with IT managers, owners, and cybersecurity professionals. Each threat was not just ranked by the severity of the threat but evaluated based on a risk exposure score (Likelihood × Impact) and placed in the cloud shared-responsibility model for IaaS, PaaS and SaaS. Results from the analysis suggest that data breaches, resource misconfiguration, and regulatory non-compliance are the top risks, while limited resources, skills gaps, and reliance on third-party providers are considered as constant constraints. The findings also reveal that the security burden is lowest for SMEs on SaaS and highest on IaaS, and that there are still several effective strategic ways to respond, including the use of multi-factor authentication, encryption and certified providers, which are still under-adopted compared to their perceived effectiveness. Finally, the study suggests a framework for risk management and some practical recommendations for SMEs and policymakers, providing both an analytical perspective to prioritize cloud security risks and actionable advice for resource-limited firms.

Keywords : cloud security, risk management, SMEs, shared responsibility, strategic responses, risk exposure.

EVALUATING THE EFFECTIVENESS OF MINING LEGISLATION IN ENHANCING OCCUPATIONAL HEALTH AND SAFETY IN KHYBER PAKHTUNKHWA, PAKISTAN

Zahir Shah — 2026-06-10

The mining sector is a critical component of Pakistan’s economy and serves as an important source of employment in the province of Khyber Pakhtunkhwa (KP). However, the industry continues to face serious occupational health and safety (OHS) challenges due to inadequate enforcement of mining regulations, outdated mining practices, limited technological adoption, and insufficient regulatory oversight. This study evaluates the effectiveness of mining legislation and its implementation in improving occupational safety and health within the mining sector of KP, Pakistan. A mixed-method research approach was adopted, combining quantitative and qualitative data collection techniques. Information was gathered through structured questionnaires, online surveys, field observations, and interviews involving 340 mine workers and 22 mine inspectors from 30 districts of Khyber Pakhtunkhwa. The collected data were analyzed using the Statistical Package for Social Sciences (SPSS). The results revealed that the mining workforce is predominantly young, with a large proportion of workers possessing limited formal education and belonging to low-income socioeconomic groups. Nearly half of the workers were found to be illiterate, while most were employed in frontline mining activities. These conditions reduce workers’ ability to understand safety instructions, regulations, and hazard warnings, thereby increasing their exposure to occupational risks. Furthermore, production-based payment systems encourage workers to prioritize output over safety, often leading to unsafe practices and non-compliance with established regulations. The study found that existing mining legislation, including the Khyber Pakhtunkhwa Mines Safety, Inspection and Regulation Act, has contributed positively to improving workplace safety and provided a useful framework for regulating mining activities. However, significant deficiencies remain in the practical implementation of these laws. A shortage of inspectors, limited field inspections, inadequate documentation of violations, insufficient training opportunities, and weak enforcement mechanisms continue to hinder effective compliance. Field observations further revealed poor use of personal protective equipment (PPE), reliance on manual mining methods, and inadequate adherence to safety standards despite the existence of regulatory requirements. The quality of legal proceedings and compensation is generally viewed as satisfactory, but the success rate of prosecutions remained relatively low, reducing the deterrent effect of regulatory actions. In addition, a substantial proportion of safety violations were not formally documented, limiting the ability of regulatory authorities to monitor trends and implement corrective measures effectively. Overall, the study concludes that while the legislative framework governing occupational health and safety in KP’s mining sector is generally adequate, its effectiveness is constrained by weaknesses in implementation and enforcement. To improve mine safety performance, the study recommends increasing the number of mine inspectors, strengthening inspection and monitoring systems, enhancing training programs for both workers and inspectors, improving accident reporting and violation-recording mechanisms, adopting modern safety technologies, and enforcing stricter legal action against non-compliant operators. Effective implementation of these measures would contribute to safer working conditions, lower accident rates, improved worker welfare, and the sustainable development of Pakistan’s mining industry.

BLOCKCHAIN-READY LEAKAGE-AWARE MACHINE LEARNING FRAMEWORK FOR SHORT-TERM SOLAR AC POWER FORECASTING AND ENERGY DATA INTEGRITY VERIFICATION

Fahad Soomro — 2026-06-15

For reliable smart-energy management, short-term (ST) photovoltaic (PV) power forecasting is crucial, as is the exchange of energy data with accurate trustworthiness. The output of PV ACs is influenced by irradiation, environment and module temperature, and short-term generation behavior, and distributed energy records require integrity verification against unauthorized modification. This paper presents a blockchain-ready and leakage-aware framework, for solar AC power forecasting of the next step and provides a verification of the integrity over the energy records. The forecasting part forecasts the next-step AC power based on the weather, temporal and lag features derived from the PV generation data and the weather sensor data. Following the pre-processing and feature engineering steps, the final data set consisted of 68,708 records from 22 inverter/source units with 54,966 records split into a training set and 13,742 records split into a test set through a chronological split. The tested regression models were: Linear Regression, Ridge Regression, Random Forest, Extra Trees, Gradient Boosting and XGBoost. To reduce direct inverter-side data leakage, in the main forecasting experiment, the power from the DC side was omitted. Extra Trees achieved the best performance with MAE = 12.8138, RMSE = 37.1821, MAPE = 3.8496%, and R² = 0.991146. A separate inverter-aware estimation experiment with DC power was retained only to demonstrate the strong electrical dependency between DC-side and AC-side PV power. For integrity verification, the best forecasting outputs were converted into hash-secured records containing plant ID, source key, timestamp, actual AC power, predicted AC power, error value, and SHA-256 hash. A total of 2,000 records were stored in the verification layer, and all 100 intentionally modified records were detected, achieving a 100% tamper detection rate. The results show that leakage-aware solar AC forecasting can be coupled with lightweight, blockchain-ready record verification in a reproducible workflow.

A HYBRID DEEP LEARNING FRAMEWORK INTEGRATING LSTM AND LIGHTGBM FOR SENTIMENT ANALYSIS OF ROMAN URDU TEXT

Kanwal Mehmood — 2026-06-15

Sentiment analysis is central to extracting opinions and emotional context from user-generated text, yet its application to Roman Urdu remains constrained by the language's informal usage, non-standardised orthography, and scarcity of annotated resources. This study proposes a hybrid classification framework that couples a Long Short-Term Memory (LSTM) network with a Light Gradient Boosting Machine (LightGBM) classifier to improve sentiment prediction for Roman Urdu. The LSTM branch models sequential and contextual dependencies in the text, while the LightGBM branch captures non-linear interactions among engineered features; the two branches are combined through a weighted Softmax fusion layer. A publicly available Roman Urdu corpus of 98,984 samples obtained from Kaggle was preprocessed using a custom tokenizer, transliteration-aware normalisation, and language-specific stop-word removal. The framework was trained and evaluated using stratified ten-fold cross-validation. The hybrid model achieved a classification accuracy of 97.74%, exceeding the standalone LSTM (93.72%) and standalone LightGBM (69.51%) models, and also outperforming conventional classifiers including Random Forest, Support Vector Machine, and k-Nearest Neighbour. The results indicate that integrating sequential representation learning with gradient-boosted feature modelling is an effective strategy for sentiment analysis in low-resource, non-standardised languages, and provide a basis for future work on multilingual and code-mixed sentiment systems.

THE ROLE OF PROMPT ENGINEERING IN LEVERAGING GENERATIVE AI FOR EARLY-STAGE STARTUPS

Muhammad Moazam — 2026-06-15

This research explores how prompt engineering can empower early-stage startups to make better use of generative artificial intelligence (AI) tools. In an era where Large Language Models like GPT-4 are becoming more deeply entrenched in startup operations, ranging from content creation to customer support, market research, and software development, the effectiveness of human-to-AI communication becomes a key factor in determining operational success. However, the majority of startup teams are not formally trained in prompt design and they have to try-and-try approaches to get these to work: sometimes they do and sometimes they don't. This study uses a quantitative pre-post comparative design with a purposive sample of 10 online-only startups to assess the improvement in the relevant indexes before and after the application of structured prompt engineering techniques in the indexes of relevance, accuracy, user satisfaction and time efficiency. The results of this study should show a significant improvement in all the measured aspects after the implementation of prompt engineering, thus proving that prompt engineering is not just a technical skill, but a strategic competency. It also outlines a recurring challenge with prompt literacy within startup teams and offers practical strategies for integrating prompt training into the onboarding and daily operations processes. Political implications related to the competitiveness of startups and the governance of AI and digital literacy education are discussed.

GREEN SOFTWARE ARCHITECTURE: CARBON-AWARE AND ENERGY-EFFICIENT APPROACHES FOR SUSTAINABLE CLOUD COMPUTING — A COMPARATIVE LITERATURE REVIEW

Kinza Noor — 2026-06-15

Cloud computing has had a tremendous impact on the energy consumption and carbon emissions of datacenter facilities around the world. Consequently, Green Software Architecture has become a significant research field, which aims at minimizing the environmental footprint while preserving the system performance. In this paper, a comparative literature review of carbon-aware and energy-efficient approaches for sustainable cloud computing is presented. Twenty research studies are examined and classified in various sustainability areas such as carbon-aware scheduling, energy-efficient resource management, renewable energy integration, AI-based optimization, and green software design. The results show that intelligent workload scheduling, renewable energy use, and machine learning optimization can significantly cut carbon emissions and energy consumption. The paper also points out the existing challenges and future research directions in the development of environmentally friendly cloud systems.

RIS-ASSISTED UAV-ENABLED CELL-FREE MASSIVE MIMO SYSTEMS FOR 6G WIRELESS COMMUNICATION

Farhan Siddiqui — 2026-06-15

The sixth-generation wireless networks are anticipated to provide ultra-reliable low latency communication, massive connectivity, high spectral efficiency, better energy performance, and seamless service availability in terrestrial, aerial and remote environments. To fulfill these requirements, recent works have been increasingly focused on integrating reconfigurable intelligent surfaces, unmanned aerial vehicles and cell-free massive multiple-input multiple-output architectures. Reconfigurable intelligent surfaces offer programmable control over the wireless propagation environment, unmanned aerial vehicles provide flexible 3D deployment and fast coverage extension, and cell-free massive MIMO improves user-centric service by coordinating distributed access points without rigid cell boundaries. This survey provides a structured overview of RIS-assisted UAV-enabled cell-free massive MIMO systems for 6G wireless communications. It covers enabling technologies, system architectures, channel modeling, UAV trajectory optimization, RIS phase configuration, resource allocation, energy and spectral efficiency, physical layer security, scalability, and practical deployment issues. The study also compares recent approaches based on performance objectives, optimization methods, application domains and implementation constraints. The review shows that the combined use of RIS, UAVs and cell-free massive MIMO can significantly improve coverage, reliability, interference management and energy-aware operation, particularly in scenarios susceptible to blockage or highly mobile or with limited infrastructure. However, imperfect channel state information, RIS hardware impairments, UAV battery constraints, synchronization overhead, high computational complexity, security risks, and lack of mature standardization still hinder the practical deployment. Based on the review of the literature, future research directions include lightweight channel estimation, AI-assisted joint optimization, energy-efficient UAV control, secure RIS configuration, multi-UAV coordination, experimental testbeds, and interoperable protocols. In summary, the RIS-assisted UAV-enabled cell-free massive MIMO is a promising architecture which is still evolving to achieve flexible, intelligent and scalable 6G wireless networks.

A TF-IDF AND LOGISTIC REGRESSION PIPELINE FOR SCHOLARLY ARTICLE CLASSIFICATION AND RECOMMENDATION: IEEE XPLORE BENCHMARK STUDY

Ghazi Irfan — 2026-06-15

The rapid growth of scholarly publications has made automated topical organization and recommendation essential for efficient literature search. However, most existing approaches treat classification and recommendation as two separate tasks with inde pendent representations. This paper proposes a unified content-based framework in which a single TF-IDF representation of the article abstract drives both multi-class topical classification and top-k article recommendation. A new benchmark of 11,744 abstracts is constructed from the IEEE Xplore digital library in six topical queries. The abstract text and the topical query label are retained for every record, so the entire pipeline operates on abstracts alone without titles, author keywords, or indexer-supplied terms. A preliminary confusion analysis reveals that two queries (Big Data Analysis and Cloud Computing) exhibit near-complete vocabulary collapse and are consolidated into a single class, yielding a five-domain benchmark: Big Data & Cloud Comput ing, Data Science, Robotics, Wireless Communication, and Breast Cancer. On the classification side, five supervised learners (Logistic Regression, Linear SVM, SGD, k-Nearest Neighbours, and Decision Tree) are compared under identical 80/20 stratified hold-out and 10-fold cross-validation protocols. The grid-searched Logistic Regression attains 85.01% accuracy (weighted F1 = 0.850), and a soft-voting ensemble of Logistic Regression, Linear SVM, and SGD reaches 85.57% (weighted F1 = 0.855). On the recommendation side, the same TF-IDF representation powers a top-10 recommender that achieves MAP@10 = 0.7664 with pure cosine ranking. Reusing the classifier’s calibrated class probabilities to re-rank cosine candidates lifts Precision@10 by +12.5 percentage points (to 0.7715) and MAP@10 by +7.6 points (to 0.8428), with consistent gains on NDCG@10 and MRR. The dataset, preprocessing pipeline, trained models, and replication scripts are released to support reproducibility.

INTERNET OF THINGS: APPLICATIONS, SECURITY, PRIVACY AND FUTURE PROSPECTS

Meerub Akhtar — 2026-06-15

The Internet of Things (IoT) is used in homes and hospitals as well as in outdoor spaces to monitor and report environmental. More useful functions. By perceiving, communicating, and acting smart in different situations, the Internet of Things (IoT) has become a major technological model of the digital age. IoT transforms traditional systems into intelligent infrastructures that enhance automation, efficiency, and decision-making across several domains, such as manufacturing, transportation, healthcare, and agriculture, by integrating sensors, embedded systems, and communication networks. The great security and privacy concerns occasioned by the enormous quantity and variety of IoT devices cannot be overstated. IoT systems are vulnerable to numerous cyber-attacks and privacy breaches based on resource depletion, weak authentication, and insecure communication protocols, and inadequate data security practices. Internet of Things' architecture, its primary applications, and the key security and privacy issues jeopardizing its reliability are comprehensively discussed in this research. It examines existing security practices such as access control models, authentication schemes, and encryption techniques while highlighting the growing role of blockchain, AI, and machine learning in the development of advanced IoT defense systems. The paper also deals with the legal and ethical implications of IoT data management and examines prospective directions for future work to build IoT frameworks that are scalable, lightweight, and privacy-preserving. The research concludes that it is crucial for building a secure and trustworthy IoT environment to have a holistic approach integrating user-centric privacy models, technological innovation, and compliance with the law.

COMPARATIVE ANALYSIS OF MACHINE LEARNING TECHNIQUES FOR NETWORK INTRUSION DETECTION IN CYBER SECURITY WITH A DIVERSE METRIC-BASED PERFORMANCE ASSESSMENT

Farhan Tariq — 2026-06-15

In modern communication and networking, the safe and reliable transfer of data is a necessity of time because the number of intruder attacks on computer networks aims to gain access to crucial information. To protect the network data from any malicious attack, the network intrusion detection systems (NIDSs) play the most critical role. It analyzes the data pattern and secures the network from any attack. This pattern analysis is not possible manually due to the large scale of data; however, machine learning (ML) is a powerful technique to analyze the large scale of data patterns and detect any malicious threats. In this work, we integrated ML with NIDS to analyze and monitor the networking data. We have applied six supervised ML techniques, which include Random, Hoeffding, and Decision Tree, Averaged One-Dependence Estimators, Instance-based KNN, and Naive Bayes, during the experiment and also considered six performance assessment criteria, which include accuracy, precision, true and false positive rates, Matthew correlation coefficient, and receiver operating characteristic area for the three different datasets. The Pareto principle is considered for the training and testing data. According to the results, A1DE is the best model among the applied techniques; it identifies patterns in the data with 99.9964% accuracy, which establishes a foundation for further research. The researchers use these findings as a starting point for determining which cyber-related attributes should be prioritized to create the most effective and successful NIDS.

SITE-SPECIFIC DETERMINISTIC SEISMIC HAZARD ANALYSIS OF A COMMERCIAL BUILDING IN ATTOCK CITY, PAKISTAN

Sajid Ayaz — 2026-06-16

This study presents a Site-Specific Deterministic Seismic Hazard Analysis (DSHA) for a proposed commercial building located in Mehria Town, Attock, Pakistan. The Attock region lies within a tectonically active zone influenced by the ongoing convergence of the Indian and Eurasian plates, making it vulnerable to moderate and strong seismic events. Major active fault systems in the surrounding area were identified and characterized, and shortest source to site distances were calculated using GIS and Google Earth tools. Peak Ground Acceleration (PGA) was estimated using two empirical attenuation relationships, Cornell, Banon et al (1977) [1] and Boore and Atkinson (2008) [2]. It was observed that among all the identified seismic sources, the Main Boundary Thrust (MBT) at epicentral distance 46.82 km with the maximum credible earthquake magnitude of Mw 7.6 was identified as the controlling seismic source. The highest estimated PGA for the site is 0.268g using the Cornell et al. relationship. The study results demonstrate that the study area is in moderate seismic hazard zone, and it is recommended to apply suitable seismic design measures following the Building Code of Pakistan (Seismic Provisions 2007).

MODELING CLOUD COMPUTING ADOPTION IN IT-RELATED EDUCATIONAL INSTITUTIONS: AN EMPIRICAL INVESTIGATION USING THE DIFFUSION OF INNOVATION THEORY

Attia Agha — 2026-06-16

This paper examined the factors that influence the acceptance and use of cloud-based services in education. The study used a quantitative cross-sectional methodology, collecting data from 194 IT learners from two IT universities in Hyderabad, Pakistan, via a Likert scale questionnaire. The study is theoretically based on the Diffusion of Innovation theory and used structural equation modeling in Amos to discover the influencing factors for adoption. The path analysis revealed that Cloud Computing Adoption is positively and significantly influenced by Compatibility (β = 0.227, CR = 2.392, p = 0.017) and Relative Advantage (β = 0.402, CR = 4.262, p < 0.001). and Observability (β = 0.11, CR = 1.964, p = 0.049). Meanwhile, contrary to the original DOI assumption, complexity had a positive and significant impact on cloud computing adoption (β = 0.399, CR = 4.115, p < 0.001). Furthermore, the relationship from Trialability to Cloud Computing Adoption was negative and non-significant (β = -0.07, CR = -0.714, p = 0.475). The study contributes to the field by investigating the cloud computing adoption factors in education from the perspectives of developing countries.

A DATA-DRIVEN APPROACH TO MONTHLY TEMPERATURE FORECASTING FOR CLIMATE ADAPTATION AND URBAN PLANNING IN KARACHI, PAKISTAN

Hira Ashraf Baig — 2026-06-16

Temperature prediction is useful in combating the constantly changing climate conditions in the urban regions with reference to aspects such as agriculture, urban development and safety. The present study aims at providing accurate predictions for the monthly average temperature of Karachi city in Pakistan using machine learning algorithms with the goal of producing robust prediction resources for climate change planning. Karachi faces challenges such as rising temperatures, the urban heat island effect, and forecasting limitations. The city needs accurate temperature data to save its assets and people from climate change. The model was checked by comparing the estimated temperature for the year 2024 with the observed values. According to the results, the 2024 predictions achieved a low Mean Squared Error of 0.49, demonstrating the high accuracy of the predictive model. For instance, the mean predicted temperature for the Karachi for May 2024 was 35.7 °C while the actual temperature was 35.8 °C, the difference of only 0.1 °C. Furthermore, the study makes two predictions and controls up to the first three months of the year 2025. The model successfully forecasted the temperatures for January, February, and March 2025, with observed average temperatures of 26.8°C for January and February, and 27.1°C for March which corresponds to the usual working season temperature patterns and validates the proposed model for long term forecasting. This investigation is helpful for reflecting Karachi’s temperature trends and will be useful for creating more efficient structures as well as preventing measures for climate change. This research helps in understanding the temperatures in Karachi effectively and has a potential for using machine learning methods to resolve environmental problems. This research highlights the potential of data-driven approaches for enhancing climate resilience and offers a practical framework for temperature forecasting in regions to support sustainable city planning.

SENTIMENT ANALYSIS OF SOCIAL MEDIA DATA FOR PAKISTANI FASHION BRAND MONITORING USING MACHINE LEARNING

Abu Horrara — 2026-06-16

The present paper is introducing an industrial grade product, called Brand Pulse, that integrates brand monitoring with the trend intelligence in the environment of the Pakistani fashion industry, inspired by the use of social media. In this paper, it is used a unique bilingual lexicon in English and a set of romanized Urdu with the help of Random forest to make binary trend direction predictions of brand trends (UP/DOWN). A total of 10,602 data points were collected from seven different platforms (Instagram, Facebook, Twitter-X, TikTok, YouTube, Daraz.pk, Google Reviews) of 17 of the top fashion brands in Pakistan. With five features (restricted to Brand ID, Platform ID, Likes, Shares and Sentiment Score), a random forest classifier model with 250 estimators and maximum depth of 3 was able to get 94% accuracy on the "free" test sample and 93-95% accuracy in each of the validation folds. The entire prediction process is available via a FastAPI RESTful API service, and an interactive Streamlit application.

CONSTRUCTION OF ELLIPTIC CURVES BASED SUBSTITUTION BOX WITH APPLICATIONS IN THE TEXT DATA ENCRYPTION

Razia Riaz — 2026-06-16

Today, the design of secure substitution boxes (S-boxes) is a crucial issue in cryptography, especially considering the sophistication of the cryptanalytic attacks. In this study, a parameterized key-dependent Mordell elliptic curve construction approach to S-box is proposed over using irreducible polynomials. Secret key is used to create key-dependent elliptic curves, adding extra randomness and creating even more security for encryption. The proposed method utilizes the algebraic properties of Mordell elliptic curve and the efficiency of the computation in finite fields to generate powerful S-boxes. The effectiveness of the generated S-box when it comes to the cryptographic properties is analyzed with some standard metrics such as nonlinearity, Strict Avalanche Criterion (SAC), Differential approximation Probability (DAP), Bit Independence Criterion (BIC), and Linear Approximation Probability (LAP). Moreover, the Avalanche effect analysis is performed for evaluating the effectiveness of encryption scheme. Analyses results showed excellent resistance to both differential and linear cryptanalysis, which demonstrates that the proposed dynamic S-box is an efficient component for modern cryptographic applications.

INTEGRATION OF BIOCHAR, PRECISION AGRICULTURE, AND GENOMIC TECHNOLOGIES FOR CLIMATE-RESILIENT CROP PRODUCTION IN PAKISTAN

Anam Iftikhar Dar — 2026-06-17

Climate change poses significant challenges to agricultural productivity and food security in Pakistan through increasing temperatures, erratic precipitation patterns, water scarcity, soil degradation, and the growing frequency of extreme climatic events. Developing climate-resilient agricultural systems has therefore become a strategic priority for ensuring sustainable crop production and environmental sustainability. This study investigated the integrated effects of biochar application, precision agriculture technologies, and genomic technologies on climate-resilient crop production in Pakistan. Grounded in the Climate-Smart Agriculture (CSA) Theory, the study proposed an integrated conceptual framework that examined the synergistic contributions of biological, digital, and genomic innovations toward enhancing agricultural resilience. A quantitative, explanatory, and cross-sectional research design was employed, and primary data were collected from 400 agricultural stakeholders, including farmers, agricultural scientists, extension officers, and researchers across Pakistan. Data were analyzed using Structural Equation Modeling (SEM). The findings revealed that biochar application, precision agriculture technologies, and genomic technologies each exerted significant positive effects on climate-resilient crop production. Moreover, their integrated adoption demonstrated the strongest influence on agricultural resilience, indicating that technological complementarities substantially improve soil health, resource-use efficiency, crop stress tolerance, and sustainable productivity. The study contributes to the growing literature on climate-smart agriculture by developing and validating a multidisciplinary framework for climate-resilient crop production. The findings provide important theoretical, practical, and policy insights for promoting sustainable agricultural intensification, strengthening food security, and enhancing climate adaptation strategies in Pakistan.

AN AI-DRIVEN BLOCKCHAIN-BASED CYBERSECURITY FRAMEWORK FOR SECURE CLOUD COMPUTING ENVIRONMENTS

Ahmed Wali Khan — 2026-06-17

Cloud computing has emerged as a foundational technology for modern digital infrastructure due to its scalability, flexibility, and cost-efficiency. However, the increasing adoption of cloud platforms has introduced significant cybersecurity challenges, including unauthorized access, data breaches, Distributed Denial-of-Service (DDoS) attacks, spoofing, insider threats, and data tampering. Traditional cloud security mechanisms suffer from centralized vulnerabilities, limited scalability, and inadequate real-time attack detection. To address these limitations, this paper proposes an AI-Driven Blockchain-Based Cybersecurity Framework (AIBCF) for secure cloud computing environments. The proposed framework integrates blockchain technology with a hybrid Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM) model to provide decentralized trust management, intelligent intrusion detection, and adaptive threat mitigation. Blockchain ensures secure authentication, immutable transaction logging, and smart contract-based enforcement, while the CNN-LSTM model performs real-time cyberattack detection and classification. Experimental evaluation on the CICIDS2017 dataset under DDoS, spoofing, brute force, and infiltration scenarios achieved 98.2% accuracy, 97.6% precision, 97.1% recall, and 97.3% F1-score, with a false positive rate of 1.8%, outperforming existing machine learning and blockchain-based baselines. Ten-fold cross-validation confirmed stable results (accuracy: 98.2% ± 0.4%). The findings indicate that integrating blockchain with AI-driven mechanisms significantly improves cloud security, reliability, and adaptive defense capabilities.

TOWARDS SAFER BATTERIES SOLID-STATE ELECTROLYTES AND INTERFACE STABILIZATION MECHANISMS

Haleema Bibi — 2026-06-17

Background: Solid-state batteries are being explored as safer options than traditional lithium-ion batteries due to the increased safety from the solid-state electrolytes which reduce risks associated with thermal runaway, leakage and flammability. But they are limited in practical performance by the instability of the electrode/electrolyte interfaces, the resistance at the interfaces, the formation of dendrites and chemo-mechanical degradation during cycling.

Objective: This work was motivated by the desire to consider the role of solid-state electrolytes and interface stabilization mechanisms for safer and longer lasting battery systems.

Method: The method used is literature based, which involves searching for studies in recent years and selecting those published in 2021-2026 that are peer-reviewed. The review was mainly concerned with oxide, sulfide, polymer, composite and quasi-solid electrolytes, highlighting the following areas: ionic conductivity, electrochemical stability, area-specific resistance, lithium dendrite suppression, artificial interphases, and cathode/electrolyte compatibility.

Result: The results indicated that oxide electrolytes results in thermal/mechanical stability, sulfide electrolytes results in high ionic conductivity, polymer electrolytes results in flexibility and composite systems results in a balance of conductivity and interfacial contact. However, the quality of the interface rather than the type of electrolyte was the major factor for safety and performance. Electron-blocking interlayers, lithiophilic coatings, cathode protective layers, molecular anchoring, entropy-stabilized interfaces and dynamically adaptive interphases decreased interfacial degradation, ensured uniform Li+ flux and inhibited dendrite growth and enhanced cycling stability.

Conclusion: Solid-state electrolytes are a potential pathway to safe, high-energy batteries, but scalable, stable and mechanically adaptable interfaces are needed for commercialization. Good engineering of the interfaces will be key to minimizing short-circuit, durability, and to implement lithium metal solid-state battery applications.

A COMPUTER VISION-BASED FRAMEWORK FOR CO-INFECTION DETECTION AND SEVERITY ASSESSMENT IN PLANT LEAVES

Abdullah Danish — 2026-06-17

Accurate quantification of plant disease severity is critical for early intervention and sustainable crop management. However, it is a challenging task due to the frequent co-occurrence of multiple pathologies on a single leaf, varying illumination conditions, and high interclass similarity among severity levels. In this paper, we present a hybrid feature representation framework for the simultaneous quantification of individual disease severity levels on a single leaf. It combines handcrafted texture descriptors with deep transformer-based visual features for robust multi-label severity analysis. Specifically, the Weighted Local Binary Pattern (WLBP) and Haralick texture features capture fine-grained local lesion variations and second-order statistical spatial relationships, while the EVA02 Vision Transformer models the global semantic context and long-range dependencies across the leaf surface. The extracted features are normalized and fused into a unified and discriminative representation. The model can estimate the exact percentage and severity grade for each identified disease. The framework was tested using images of cherry and pear leaves from the PlantCity dataset, which show complex symptomatic patterns in stone and pome fruits. Experimental results show that the proposed fusion strategy is able to achieve higher classification accuracy of 84.14% for cherry and 85.42% for pear leaves, able to classify signatures of disease conditions with overlapping features successfully and better than the individual feature extractors. The results demonstrate that the integration of these global features with local features extracted using texture descriptors greatly enhances the granularity of disease classification and ensures a reliable approach for accurate multi-symptom diagnosis in smart farming applications.

6G-ENABLED AI-BASED SENSING AND COMMUNICATION CONVERGENCE: A COMPREHENSIVE SURVEY

Saqib Islam — 2026-06-17

The sixth generation of wireless networks is seeking to achieve the vision of integrated sensing and communication (ISAC), where wireless systems not only transmit data but also sense the environment. This concept allows for high-resolution mapping of environments, autonomous navigation, and extended reality that is immersive. However, implementing ISAC in 6G networks faces major challenges such as spectrum coexistence, shared hardware, adaptive waveform design, and real-time adaptability in dynamic environments. ISAC-enabled AI and ML applications for intelligent resource allocation, robust channel estimation, adaptive beamforming, and ISAC security fortification make them critical for ISAC. This survey offers a thorough overview on the inclusive backbone technologies like THz communications along with massive multiple input multiple output (MIMO) systems and reconfigurable intelligent surfaces (RIS), as well as the current approaches to ISAC channel modeling such as stochastic, deterministic, and hybrid models. Unique focus is given to AI techniques and deep learning, reinforcement learning, privacy-preserving federated ML and the issues of security and the interventions. This survey looks into practical use cases, existing models, as well as gaps in the research and is intended to be the starting point toward the development of AI-driven ISAC for 6G networks.

RECONCILING PRIVACY, EXPLAINABILITY, AND FEDERATED LEARNING IN DECENTRALIZED PRECISION AGRICULTURE

Muhammad Owais — 2026-06-17

Precision agriculture is decentralized, relying on heterogeneous datasets provided by farms, cooperatives, and sensing platforms; nevertheless, these datasets cannot be centrally aggregated owing to privacy, regulatory, and economic limitations. Although deep learning has achieved significant performance increases in crop monitoring and disease detection, it not only uses centralized training, which conflicts with the distributed nature of agricultural data, but also leads to issues of confidentiality, accountability, and trust. Previous studies have mostly studied federated learning, privacy protection, and explainable artificial intelligence individually, with no evaluations of their complexity in situations that are defined by non-IID data, sparse connectivity, and edge-computational limitations. We provide an analytical synthesis in this study, where we consider federated learning, privacy, and explainability as design requirements that are mutually dependent. We propose a common taxonomy of architectures, federated learning frameworks, privacy preservation methods, explainability methods, data formats, and deployment platforms. Through comparative analysis, we reveal trade-offs between predictive accuracy, interpretability fidelity, communication overhead, and privacy robustness, and new challenges, especially the instability of explanations, lack of auditable and decentralized benchmarks, and trade-off between privacy and utility.

OPTIMIZING DATA TRANSMISSION IN CLUSTERED MULTI-EDGE COMPUTING FOR INTELLIGENT IOT

Zain Ul Abedeen — 2026-06-17

This paper focuses on the challenges of optimizing data transmission in clustered Multi-Access Edge Computing (MEC) systems for Internet of Things (IoT) applications. With all of this proliferation of IoT devices, the traditional cloud-based architectures are limited by means of latency, bandwidth and energy efficiency. To address these challenges, this work introduces a novel data transmission optimization model which leverages dynamic clustering, reinforcement learning based task offloading, and adaptive routing approaches to optimize system performance. The proposed model aims to minimize end-to-end latency, energy consumption, and maximize throughput and packet delivery ratio (PDR) in a large-scale IoT environment. To assess the effectiveness of the proposed model the simulations were carried out with respect to static clustering and threshold offloading baseline models. The results validate the superiority of the proposed system with respect to the key performance metrics compared to the baseline systems. In particular, the proposed model achieved up to 40% less latency, 31-35% better energy efficiency, as well as a higher PDR and throughput than static clustering and threshold offloading. Furthermore, the proposed system exhibited cluster stability for 120 minutes which is much larger than that of baseline models (75-90 minutes). Moreover, the sensitivity analysis indicated that the proposed model is scalable and adaptable and works well in different node density and traffic loads. The results demonstrate the promise of MEC for making large-scale IoT networks energy efficient, low latency, and efficient. The findings of this research could help to optimize the data transmission in MEC-based IoT systems, which have potential applications in smart cities, healthcare, and industrial automation fields.

PHARMA-CHAIN: A BLOCKCHAIN-ENABLED, IOT-POWERED SUPPLY-CHAIN TRACEABILITY FRAMEWORK ON HYPERLEDGER FABRIC FOR COMBATING COUNTERFEIT AND SUBSTANDARD MEDICINES

Muhammad Usman — 2026-06-17

Background: Substandard and falsified (SF) medicines are a persistent global health emergency that is disproportionately concentrated in low- and middle-income countries (LMICs). The World Health Organization estimates that approximately one in ten medical products in LMICs fails quality testing, and pooled meta-analytic evidence places the prevalence at 13.6%, with antimalarials and antibiotics most heavily affected. Conventional pharmaceutical supply chains rely on fragmented, centrally held, paper-based or siloed digital records that are easy to forge, difficult to audit, and slow to mobilise during recalls, creating fertile conditions for counterfeit penetration.

Objectives: This study designs, models and evaluates Pharma-Chain, a permissioned blockchain and Internet-of-Things (IoT) traceability platform built on Hyperledger Fabric — that delivers immutable, end-to-end provenance of every drug batch from manufacturer to patient, enables instantaneous QR-based authenticity verification, and provides the Drug Regulatory Authority of Pakistan (DRAP) with real-time oversight, recall and quarantine capabilities.

Methods: We adopted a design science methodology. System requirements were captured through a Unified Modelling Language (UML) use-case model spanning five actors and six functional packages; interaction logic was specified through a sequence diagram tracing a transaction from the React front end through a Node.js gateway to Fabric chaincode, the Raft ordering service, and a CouchDB-backed world state; and the full business process was formalised as a swim-lane activity diagram. Four chaincodes (manufacturing, transfer, retail, and recall) were implemented and benchmarked for throughput, latency, and authentication accuracy under increasing transaction loads.

Results: The prototype sustained a committed throughput of up to 471 transactions per second (TPS) before saturation, maintained sub-second confirmation latency below 500 TPS, and executed read-only verification queries in under 0.20 s. Across four field verification scenarios, the system correctly authenticated genuine batches in 98.7% of cases, flagged 100% of counterfeit/unknown QR codes, and recalled batches. Relative to a conventional baseline, modelled supply-chain capability improved by a factor of two to three across traceability, tamper-resistance, recall speed and counterfeit detection.

Conclusions: A permissioned, IoT-integrated blockchain is a technically viable and operationally compelling instrument for securing the pharmaceutical supply chain in resource-constrained settings. Pharma-Chain aligns with international serialization regimes (DSCSA, EU FMD) while remaining tailored to Pakistan's governance realities, offering a deployable blueprint for a national drug-authentication infrastructure.

LOSS FUNCTION ANALYSIS FOR CLASS-IMBALANCED MULTI-ORGAN SEGMENTATION OF THE GASTROINTESTINAL TRACT IN MRI

Moavia Hassan — 2026-06-17

MRI guided radiotherapy for the abdominal cancers should must be marked on every scan slice to stomach, small bowel and large bowel so the radiation can avoid healthy tissues. It is often marked by hand, and different experts often outline the same organ differently. While the Deep learning can perform this task automatically, but the data makes it hard to accurate marking. Such as UW-Madison gastrointestinal (GI) tract dataset almost contains 57% no organ and remaining covers only a portion of image when organ appears that leaves the classes heavily imbalanced. The training loss is the main mechanism that drives a network to attend to such rare foreground, yet it is usually chosen by convention rather than by evidence. We compare five losses under identical conditions on a fixed 2.5D network that pairs a SegFormer MiT-B2 encoder with a U-Net decoder: Dice, soft binary cross-entropy (SoftBCE), their combination, Tversky, and a Focal-Dice combination. Training and evaluation use a patient-grouped split and per-image-averaged Dice, intersection over union (IoU), sensitivity, specificity, and precision. All five reach comparable overall Dice within 0.007 (0.9006 to 0.9072), so overall accuracy is largely insensitive to the loss here. The error profile differs sharply, however: Tversky gives the highest sensitivity (0.9465) at the lowest precision (0.9091), SoftBCE the highest precision (0.9363) at the lowest sensitivity (0.9255), and Focal-Dice the best balanced Dice (0.9072). The small bowel stays hardest under every loss. The loss should therefore be chosen for the clinically preferred balance between missing tissue and over-contouring, not for overall accuracy.

ECO-FRIENDLY FABRICATION OF MANGANESE–NICKEL–VANADIUM SULFIDE-BASED COMPOSITES WITH ZNO, TIO₂, AND AG FOR ENHANCED SUPERCAPACITOR PERFORMANCE

Nida Afzal — 2026-06-15

We report a scalable, environmentally benign hydrothermal approach for synthesizing ternary manganese–nickel–vanadium sulfide (Mn–Ni–V–S) composites decorated with zinc oxide (ZnO), titanium dioxide (TiO₂), and silver nanoparticles (Ag NPs) for high-performance supercapacitor electrodes. The entire fabrication protocol employs water as the primary solvent and avoids hazardous organic precursors, rendering the synthesis green and sustainable. Structural characterization by X-ray diffraction (XRD), Raman spectroscopy, and high-resolution TEM confirms phase-pure sulfide nanostructures with intimate interfacial coupling. The BET surface area reaches 318.4 m² g⁻¹, indicating a highly porous architecture. Electrochemical evaluation in 2 M KOH reveals a specific capacitance of 1872 F g⁻¹ at 1 A g⁻¹, energy density of 58.6 Wh kg⁻¹, and 93.7% capacitance retention over 10,000 cycles. The synergistic contributions from multiple redox-active sulfide phases, Ag-mediated conductivity enhancement, and ZnO/TiO₂ heterojunctions collectively amplify the pseudocapacitive response. This work establishes a green-chemistry pathway toward next-generation energy storage materials.

ENHANCED ROUTE VALIDATION MECHANISM TO MITIGATE THREE-NODE INSTABILITY IN ROUTING INFORMATION PROTOCOL

Younas Iqbal — 2026-06-18

The three-node instability problem of the Routing Information Protocol (RIP) is examined in this study as one of the manifestations of the count-to-infinity problem in distance vector routing protocol, which is crucial in this context. Although the split-horizon as well as the poison reverse are effective in the case of two node instability causes by the loops, they do not stop routing loops in case of three nodes and thus delay convergence and deteriorate the performance of the network. In order to counter this shortcoming, the solution of verification is suggested, according to which routers verify alternative routes with the original source before accepting them. This helps to avoid the spread of the outdated or misleading updates and provides stable routing decisions. The method proposed is demonstrated with the help of a detailed example based on Forouzan Data Communications and Networking with the flowcharts, pseudo-code, and graphical simulation. After comparative analysis, it can be seen that, verification-based method has a higher convergence rate, ensures that loops are avoided, and is more stable than simple distance vector routing and split horizon with poison reverse. The results identify the efficiency and feasibility of the presented solution, and further effort recommends the expansion of the mechanism to bigger topologies, incorporation of the newest protocols and the use of intelligent algorithms to enable proactive loop recognition.

AN INTELLIGENT TASK SCHEDULING APPROACH FOR FOG COMPUTING

Tuba Younas — 2026-06-18

By extending cloud computing to the network's edge, fog computing is a distributed computing paradigm that makes it possible to handle and analyze data in real time closer to its source. However, efficient task scheduling is necessary in fog computing optimize performance indicators such as latency, power consumption, and resource utilization. To overcome these difficulties, this study suggests Dynamic Scheduling Technique for Real-time Applications (DSTRA) based on reinforcement learning methods. The goal of the technique is to enhance the overall performance of fog computing systems by lowering latency and power consumption. Using real-time feedback from the fog nodes, DSTRA uses reinforcement learning to dynamically modify task priorities and resource allocation. With this strategy, the system can adjust to shifting circumstances and make the best scheduling choices possible in a dynamic environment. To ensure that latency-sensitive applications receive the necessary resources, tasks are prioritized based on their importance and deadline constraints. The DSTAR algorithm is evaluated through extensive simulations and real-world deployments, showing a 90% to 98% improvement in efficiency across key metrics including latency, power consumption, and overall system performance when compared to traditional scheduling approaches. This study addresses the critical resource needs of latency-sensitive applications by proposing a task-prioritization framework focused on importance and deadline constraints. We introduce the DSTRA algorithm, a robust solution for managing heterogeneous parallel task flows under dynamic constraints. DSTRA significantly outperforms conventional scheduling strategies. System delays are reduced by 90% to 98% Marked improvements are observed in power consumption and energy management, Overall resource allocation efficiency and system performance are substantially enhanced. The results confirm DSTAR’s efficacy in navigating complex, uncertain environments while maintaining optimal operational throughput.

ARTIFICIAL INTELLIGENCE FOR SUSTAINABLE SMART CITIES: MACHINE LEARNING APPLICATIONS IN INTELLIGENT URBAN GOVERNANCE AND INFRASTRUCTURE

Rehan Ali Khan — 2026-06-19

The rapid expansion of urban populations and increasing sustainability challenges created a growing demand for intelligent technologies capable of improving governance effectiveness and infrastructure performance. This study examined the role of artificial intelligence and machine learning applications in advancing sustainable smart cities through intelligent urban governance and infrastructure management. A quantitative research design was employed to investigate the perceptions of professionals involved in smart city initiatives. Data were collected from a sample of 350 respondents, including urban planners, municipal administrators, infrastructure managers, policymakers, and technology specialists. The study utilized descriptive statistics, reliability analysis, and one-sample t-test analysis to evaluate the research objectives. The findings revealed strong support for the adoption of intelligent technologies in urban environments. Artificial Intelligence Adoption recorded a mean score of 4.19 with a standard deviation of 0.61, while Machine Learning Applications achieved a mean score of 4.24 with a standard deviation of 0.58. Intelligent Urban Governance reported a mean value of 4.15 and Sustainable Smart City Development achieved the highest mean value of 4.28. Reliability analysis indicated strong internal consistency, with Cronbach’s Alpha coefficients ranging from 0.835 to 0.879, while the overall reliability coefficient reached 0.856. Furthermore, one-sample t-test results demonstrated statistically significant support for all study variables at p = 0.000. The study concluded that artificial intelligence and machine learning technologies enhanced governance responsiveness, improved infrastructure efficiency, strengthened resource management practices, and supported sustainable urban development. The findings provided valuable insights for policymakers, urban planners, and technology developers seeking to create resilient, efficient, and environmentally sustainable smart cities.

SMART TECHNOLOGIES FOR WATER SEWAGE SYSTEMS AND DECISION-MAKING WITH CIRCULAR SPHERICAL FUZZY FRAMEWORK

Muhammad Ahmad — 2026-06-17

This work aims to improve decision-making (DM) processes by utilizing the circular spherical fuzzy set (Cr-SFS), a flexible structure for managing uncertain human opinions. This paper presents a new class of AOs, such as the circular spherical fuzzy Dombi weighted averaging (Cr-SFDWA), Circular spherical fuzzy Dombi weighted geometric (Cr-SFDWG), circular spherical fuzzy Dombi order weighted average (Cr-SFDOWA) and circular spherical fuzzy Dombi order weighted geometric (Cr-SFDOWG) operators which are specially designed for Cr-SF information systems. These operators' realistic qualities and exceptional cases are clarified, emphasizing how well they fit into real-world situations. A novel methodology for MADM is applied to various real-world applications with varying needs or features. An example of an AI selection process in a water sewage system is provided to show the effectiveness of the suggested methodologies. Moreover, a comprehensive comparison method is presented to illustrate the effectiveness and relevance of proposed aggregation strategies by comparing their outcomes with those of the existing approaches. The study is accomplished with a summary of its findings and a discussion of its prospects as we advance, highlighting the potential contribution of the suggested research to the advancement of decision-making techniques in dynamic and complex environments.

VOICE CONTROLLED AND TFT TOUCHSCREEN CONTROLLED WHEELCHAIR

Engr. Aymen Jamil — 2026-06-18

In today’s world, widespread prevalence of lost limbs and sensing system is of major concern in present day due to accident, age and health problems. To assist people with such defects, the proposed intelligent wheelchair system is used which have dual control for navigation in familiar environments. This paper is related to voice command and touchscreen display based model of a wheelchair. The smart wheelchair system used the voice recognition module V3 and a 2.8” TFT Touchscreen display. Wheelchair is facilitating the movement of people who are disabled or handicapped and elderly people. The wheelchair design will allow people to do their basic daily tasks without any dependence on other person. In building the circuit for this project, we are using AURDINO MEGA and its interfacing with TFT Touchscreen module and voice recognition module with direct current motors for movement of wheelchair in different directions. The system has been designed and implemented in a cost-effective way so that if our project is commercialized the needy users in developing countries will benefit from it.

AGENTIC AI-BASED INTELLIGENT STUDY ASSISTANT USING LL MS AND VECTOR DATABASES

Zaviyar Hasnain Bhutta — 2026-06-16

By introducing intelligent and automated learning solutions, artificial intelligence (AI) has transformed contemporary educational systems. An Agentic AI-Based Intelligent Study Assistant is presented in this study. It makes use of Vector Databases and Large Language Models (LL Ms) to provide students with intelligent, context-aware, and individualized academic assistance. Natural language interaction is used to answer questions, summarize study materials, make notes, and help students learn more quickly with the proposed system. The primary focuses of the research are the creation and implementation of an intelligent system that is able to comprehend user input, retrieve relevant information through vector-based semantic search, and generate precise responses through advanced AI models. By combining the capabilities of language generation and external knowledge retrieval, the integration of Retrieval-Augmented Generation (RAG) techniques enhances the relevance and quality of responses. The system architecture includes components such as user interface, Agentic workflow, embedding models, vector database, and LLM integration. The model that has been proposed aims to make education more adaptable, to make learning easier, and to make students more productive. According to experimental analysis, the intelligent assistant performs better than conventional keyword-based systems in terms of response accuracy, contextual understanding, and user interaction. This study demonstrates how intelligent assistants can support contemporary learning environments through automation, personalization, and effective knowledge retrieval, as well as the growing role Agentic AI systems are playing in education.