A COMPARATIVE MACHINE LEARNING APPROACH TO DIABETES PREDICTION: INTEGRATING BEST FIRST SEARCH FEATURE SELECTION WITH SVM AND NAÏVE BAYES CLASSIFIERS

Abstract

Data mining is the process of using machine learning, statistical, and other database systems techniques to extract useful, hidden patterns from large data sets. It is beneficial for decision-making and prediction tasks across various domains. In the healthcare domain, data mining has been recognized as crucial for analyzing patient data to enhance the precision of patient diagnosis, provide best practices in treatment, and improve the ability to predict disease. Recent advances in   Artificial Intelligence (AI) and Machine Learning (ML) have shown remarkable potential to transform diabetic data by enabling predictive, personalized, and data-driven healthcare systems.  This study presents a comprehensive comparative analysis of multiple AI-driven models for the   robust early prediction and clinical evaluation of Diabetes. By utilizing the Best First Search algorithm for feature selection in conjunction with Support Vector Machine and Naïve Bayes classifiers to identify the most significant attributes in the dataset, our study seeks to enhance predictive performance. With an accuracy of 91.34%, our analysis showed that the SVM (Support Vector Machine) with Best First Search performed better than the other classifiers, while the Naive Bayes classifier with Best-First Search produced a reasonable result, with an accuracy of 87.88%. To improve the diabetes prediction set for early diagnosis and health management, the results of this study demonstrate the effective application of combining feature selection with classification.