<![CDATA[Maternal mortality rates are a significant global health challenge, especially in developing countries that lack adequate medical resources. Early detection of pregnancy risks is crucial to prevent serious complications. This study developed a predictive model for pregnancy risks using four machine learning techniques: Support Vector Machine (SVM), Random Forest, XGBoost, and Gaussian Naive Bayes. The data set, taken from sources such as Kaggle and the UCI Repository, includes seven physiological indicators. Data preprocessing includes data cleaning, feature normalization, and label encoding. To address class imbalance in the Low Risk, Mid Risk, and High Risk categories, the Synthetic Minority Oversampling (SMOTE) technique is used. Model evaluation used metrics such as accuracy, precision, recall, F1 score, ROC-AUC, and processing time. The results showed XGBoost as the best model, with an accuracy of 0.8566, precision of 0.8567, recall of 0.8566, F1 score of 0.8554, and ROC-AUC of 0.9633. Random Forest produced comparable results, while Gaussian Naive Bayes was the fastest but least effective. The use of SMOTE with various metrics improved the model's ability to identify high- risk cases. Ultimately, XGBoost and Random Forest are recommended for integration into medical decision support systems aimed at early detection of pregnancy risks.]]>
Copyrights © 2026
