The Effect of Resampling Techniques on Model Performance Classification of Maternal Health Risks

Indonesia's maternal mortality rate was the second highest in ASEAN, reflecting the problem of class imbalance in maternal health data. This research aimed to improve prediction accuracy in the classification of pregnant women's diseases through the application of various resampling methods. The methods used in this research included Synthetic Minority Over-sampling Technique (SMOTE), SMOTE-Edited Nearest Neighbor (SMOTE-ENN), Adaptive Synthetic Sampling (ADASYN), and ADASYN-ENN, using five classification algorithms: Decision Tree, K-Nearest Neighbor (KNN), Naïve Bayes, Random Forest, and Support Vector Machine (SVM). Performance evaluation was carried out using accuracy, precision, recall, and F1-score metrics to determine the best method and algorithm. The results showed that the SMOTE-ENN and ADASYN-ENN methods significantly improved the model's performance in predicting maternal disease. Random Forest and Decision Tree algorithms showed the best results in terms of accuracy and consistency. These findings provided practical guidance for the application of resampling techniques in the classification of pregnant women's health data, which could contribute to improving the quality of maternal health services in Indonesia.