A random forest algorithm for high-risk pregnancies prediction based on explainable artificial intelligence (XAI)

Abstract

This study explores using the Random Forest (RF) algorithm with the Explainable Artificial Intelligence (XAI) approach to predict High risk in pregnant women. Preeclampsia is a disorder that occurs during pregnancy and is characterized by hypertension and organ damage. If not treated early, it can endanger both the mother and fetus. The RF algorithm was chosen because it can process complex data, resist overfitting, and achieve good classification performance. This study used 299 data points from pregnant women at the Community Health Centre Anak Air in Padang City. The SHAP (Shapley Additive Explanations) and LIME (Local Interpretable Model-agnostic Explanations) methods were then applied to determine the role of each feature in the prediction process. The results of the analysis show that the RF model effectively identifies pregnancy risk factors, particularly preeclampsia, such as age, body temperature, diastolic blood pressure (BP), blood sugar (BS), heart rate, and urine protein. The model achieved an accuracy of 78%, meaning that 78% of its predictions align with the actual data. The model was interpreted using SHAP and LIME. SHAP reveals globally important features, while LIME provides local explanations based on individual patient characteristics. Users, especially healthcare professionals, can more clearly and informatively understand the prediction results. This approach improves system accuracy and promotes clinical acceptance of AI-based predictive technology. The system can support the early detection of preeclampsia and more accurate decision-making in maternal healthcare.