Domestic violence against women remains a persistent and underreported issue in the Philippines. This study applies five (5) machine learning models, namely: KNN, Logistic Regression, Random Forest, XGBoost, and LightGBM, to predict the vulnerability of Filipino women to domestic violence using data from the 2022 NDHS. These models were evaluated using comprehensive metrics, including Accuracy, Precision, Recall, F1-Score, Log Loss, MCC, ROC-AUC, PRC-AUC, and Confusion Matrix. Logistic Regression emerged as the most effective model, demonstrating a strong balance between interpretability and predictive performance, particularly in Accuracy, F1-Score, PRC-AUC, and ROC-AUC. The analysis identified key predictive variables that emerged from the best-performing model’s feature importance, namely: household composition, reproductive health knowledge, unmet need for contraception, frequency of internet usage/media exposure, asset ownership, and age of household heads. Based on these findings, several policy recommendations are proposed: expanding VAW desk services and barangay-level response systems; integrating domestic violence screening into family planning and maternal health programs; designing age-specific interventions such as economic support and education; and encouraging the ethical use of machine learning tools by social welfare agencies for early detection. Ethical considerations were also examined, including data privacy, bias mitigation, and responsible AI deployment. This study contributes to the growing field of data-driven policy making by demonstrating the value of machine learning in informing targeted and preventative social interventions.
Keywords: Domestic Violence, machine learning, predictive analysis
