Explainable Artificial Intelligence for Breast Cancer Classification Using SHAP: A Comprehensive Analysis

Breast cancer is one of the top life-threatening cancers in women around the world which itself suggests there should be an imperative need for accurate and reliable diagnostic tools. Related to the above fact, although machine learning models show excellent performance in classifying instances of breast cancer, often their "black-box" nature limits their acceptance in clinical practice. Explainable AI (XAI) therefore gives a gateway to improving this challenge by making the model predictions more transparent and interpretable, helping the clinicians to further understand and trust the results. This work introduces an interpretable machine learning framework for classifying breast cancer using (SHAP) SHapley Additive exPlanations for explaining model predictions. At first, a breast cancer dataset is analyzed using multiple models where the ExtraTreesClassifier performs with 97% accuracy. Next, the SHAP method is employed to determine the important features in the model for making it globally and locally interpretable in more understandable ways using various types of visualizations such as summary and force plots. This insight contributed much to understanding how the model makes such decisions and hence matching these conclusions with clinical reasoning. The proposed framework provides a high predictive accuracy value with improved interpretability.