Interpretable Machine Learning Model for Breast Cancer Prediction Using LIME and SHAP

Breast cancer, which occurs in both men and women, causes approximately 10 lakh deaths globally and has no specific risk factors. The time frame of the treatment is a long-drawn process based on the person, the type of cancer, and its level of spread. It is imperative to detect this cancer early on in order to prevent mortality. Given the prediction's significance, an accurate breast cancer prediction model must be developed. This study explores the Breast Cancer Prediction dataset, applies SMOTE (Synthetic Minority Over-sampling Technique) to balance the dataset, and proposes an effective Machine Learning (ML) model fused with Explainable AI to provide health professionals with explanations. ML algorithms are analyzed before and after applying Principal Component Analysis (PCA), and visualization is performed using t-SNE. ML algorithms, Support Vector Machine (SVM), k-Nearest Neighbor (kNN), Random Forest (RF), Stochastic gradient Descent, XGBoost, Gradient Boosting, Decision Tree (DT), and Naïve Bayes are trained on the Breast Cancer dataset. It is seen that the RF model outperforms other models considered with 95.9% accuracy. To understand the weightage of features considered by the best breast cancer prediction models and to provide trust to the doctors, the Explainable AI (XAI) packages LIME (Local Interpretable Model-agnostic Explanations), and SHAP (SHapley Additive exPlanations) are used. The effective ML model with XAI techniques empowers clinicians with actionable insights for more informed breast cancer diagnosis and treatment decision-making.