A Comparative Study of Machine Learning Classifiers and Ensemble Method for Breast Cancer Detection Using XAI Technique

Breast cancer remains a leading cause of mortality among women, necessitating accurate and explainable diagnostic tools. While machine learning (ML) models can enhance diagnostic accuracy, their interpretability is crucial for clinical adoption. A novel framework integrating Random Forest classifiers with ensemble techniques, SHapley Additive exPlanations (SHAP), and Local Interpretable ModelAgnostic Explanations (LIME) to provide both high performance and transparency in breast cancer diagnosis is proposed. The performance evaluation of Random Forest, Neural Networks, and feature selection-enhanced Random Forest models, with validation accuracy of 97%, 98%, and 96%, respectively. Our ensemble approach combining Random Forest (RF), Logistic Regression (LR), and SVM achieved 95% accuracy. SHAP highlights key features like menopause and tumor size, while LIME offers case-specific explanations to improve trust in AI-driven decisions. This framework delivers both robust accuracy and interpretability, making it a valuable tool for advancing ethical AI in healthcare.