An Explainable Ensemble Learning Approach for Gestational Diabetes Prediction

The application of machine learning in healthcare has become increasingly critical for the early detection and diagnosis of diseases. This paper presents a comprehensive approach to predict Gestational Diabetes Mellitus (GDM) using a variety of machine learning techniques, including Random Forest, Logistic Regression, Naive Bayes, K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Gradient Boosting. The dataset underwent extensive preprocessing, which included handling missing values through mean imputation(mean for numerical data+mode for categorical data), label encoding categorical features, and addressing class imbalance via SMOTEENN. Outliers removal was performed using the Z-Score method to enhance model performance. Ensemble techniques such as Stacking and Voting were employed to improve predictive accuracy. Model performance was evaluated using metrics including accuracy, ROC-AUC score, and confusion matrices. Additionally, SHapley Additive exPlanations (SHAP) values were utilized to provide interpretability and insights into feature importance, ensuring transparency in decision-making. The proposed approach demonstrated robust performance in predicting GDM, thereby contributing to timely interventions and better healthcare outcomes.