A Predictive Analytics System for Assessing Gestational Diabetes and Associated Maternal-Fetal Outcomes

Gestational diabetes mellitus (GDM) remains a leading metabolic complication of pregnancy that drives substan- tial maternal, perinatal, and lifelong health risks. This study developed and internally validated a clinical-decision-support system that predicts both the occurrence of GDM and its major adverse outcomes using only routine data available at the first antenatal (booking) visit. From a retrospective cohort of$\mathbf{5, 4 3 2}$pregnancies, thirty-two demographic, anthropometric, historical, and biochemical variables were processed through five machine-learning algorithms. The XGBoost model deliv- ered the highest performance with an area under the receiver operating characteristic curve of 0.892 (95% CI$0.877-0.906$), sensitivity 85.3%, specificity 87.1%, and$F 1$-score 0.81 on the independent test set- markedly superior to conventional riskfactor scoring. SHapley Additive exPlanations (SHAP) provided full individual-level and global interpretability, confirming prepregnancy BMI, previous GDM, family history, and early fasting glucose as dominant drivers. High-risk women identified by the model exhibited 3.1-4.8-fold higher odds of preeclampsia, cesarean delivery, and large-for-gestational-age infants. A web-based prototype was implemented that generates instantaneous risk probabilities and evidence-based management recommendations. By shifting accurate risk stratification from$\mathbf{1 2} \boldsymbol{-} \mathbf{1 6}$weeks earlier than current universal screening, the system enables timely lifestyle modification, pharmacotherapy when indicated, and intensified surveillance, offering a practical pathway toward precision antenatal care and interruption of the intergenerational cycle of cardiometabolic disease.