Caesarean Section Classification Using Machine Learning with SHAP Explainability

Caesarean sections (C-Section), performed for medically essential reasons, are surgical procedures that can significantly reduce maternal and neonatal mortality. Artificial Intelligence-driven Clinical Decision Support Systems (AICDSS) help doctors to make well-informed choices. However, existing ML methods for predicting C-Sections often suffer from limited generalization across diverse clinical profiles, poor interpretability, and overfitting on small, imbalanced datasets. An ML-based framework combined with an Explainable Artificial Intelligence (XAI) method namely SHapley Additive exPlanations (SHAP) has been suggested as a solution to these drawbacks in order to improve prediction performance and model transparency. Categorical Boosting (CatBoost), Histogram Gradient Boosting (HGB), Gaussian Naïve Bayes (GNB), and a Stacking ensemble classifier were trained using stratified 10-fold cross-validation on the UCI C-Section Dataset. The hyperparameters of these models are searched using the Random Search technique. SHAP values measure feature contributions, allowing for both local and global interpretability. CatBoost demonstrated strong scalability, achieving the highest average training accuracy (81.25%) with efficient CPU usage. It achieves the highest mean accuracy of 100%, recall of 100%, precision of 100%, F<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf>-score of 100%, and Geometric Mean (GM) of 100%. Consistent model explainability through SHAP, robust feature interaction learning through gradient boosting, mitigation of overfitting through SHAP-guided validation, low computational overhead for real-time applicability, and enhanced generalization across small, unbalanced clinical datasets are all provided by the suggested method. Compared to traditional black-box models, the proposed framework is well-suited for CDSS applications due to its strong performance and transparent decision-making.