Explainable AI-Based Classification of Birth Methods In Rural Areas Using Random Forest: A Model Development Study

The rise in maternal mortality rates, particularly in rural areas, underscores the urgent need for accurate and interpreted diagnostic tools to help healthcare professionals make informed decisions about delivery methods. This study proposes an Explainable Artificial Intelligence (XAI)-based classification model for predicting delivery methods (vaginal or cesarean) using the Random Forest algorithm, enhanced with Local Interpretable Model-Agnostic Explanations (LIME). The model was developed through Design Science Research methodology and trained on 300 medical records from midwifery practices in the rural Bandung Regency. Data preprocessing involved data cleaning, categorical feature encoding, and train-test splitting. Model performance optimization was achieved using hyperparameter tuning via RandomizedSearchCV. Evaluation metrics included accuracy, precision, recall, F1-score, confusion matrix analysis, and validation through 5-fold cross-validation to ensure robustness and generalizability. The results showed an accuracy of 95%, with F1-scores of 0.96 for vaginal delivery and 0.94 for cesarean section. LIME visualizations identified abnormal fetal position, prior cesarean history, and fetal macrosomia as key predictors influencing cesarean outcomes. Despite promising results, the model's generalizability might be limited by the exclusive use of six clinical features and data from a single rural region. This study shows that the model provides accurate and transparent predictions, making it useful for supporting maternal care in low-resource settings.