BREAST CANCER DETECTION USING MACHINE LEARNING AND EXPLAINABLE AI TECHNIQUES

Abstract

Breast cancer arises from the uncontrolled growth of abnormal cells within breast tissue, often progressing into malignant tumors capable of spreading and becoming life-threatening. A complex interplay between environmental influences and individual genetic makeup contributes to the development of this condition. With the increasing prevalence of breast cancer, enhancing early detection and treatment methods is critically important. In recent years, machine learning (ML) has shown significant promise in improving diagnostic accuracy in the medical field. This study explores the use of patient diagnostic features alongside various ML classifiers to identify breast cancer cases effectively. Explainable Artificial Intelligence (XAI) methods were integrated to provide interpretability and transparency into the model predictions. Among the models evaluated, the Random Forest classifier achieved the highest F1-score of 84%, while a stacked ensemble approach—combining multiple model outputs—achieved an F1-score of 83%. The study further examined explanation techniques such as SHAP (SHapley exPlanations), LIME (Local Interpretable Model-agnostic Explanations), ELI5, Anchor, and QLattice to interpret model behavior and feature importance. These interpretable approaches have the potential to support clinicians in making more accurate diagnoses, minimizing errors, and improving decision-making processes in breast cancer care