ANALISIS KINERJA MODEL STACKING BERBASIS RANDOM FOREST DAN SVM DALAM KLASIFIKASI RUMAH TANGGA BERDASARKAN GARIS KEMISKINAN MAKANAN DI PROVINSI JAWA BARAT

The stacking method is an ensemble technique in machine learning that combines predictions from several base models to improve classification accuracy. This research applies the stacking method with two machine learning algorithms, namely Random Forest and Support Vector Machine (SVM) as base learners and logistic regression as a meta learner. This study aims to develop a classification model to identify households based on the food poverty line in West Java Province. The data used is KOR and household data in West Java Province sourced from the 2023 BPS National Socio-Economic Survey (Susenas). The variables used consisted of 24 independent variables with food poverty level as the response variable. Modeling was conducted using feature selection using Recursive Feature Elimination (RFE) and class imbalance handling using the ADASYN method. The results showed that the stacking model was superior to the single model with a balance accuracy of 0.81, sensitivity of 0.72, and specificity of 0.89. Feature importance analysis identified that calorie consumption, expenditure on cigarettes, meat and fruits, and expenditure on rice, eggs and other commodities contributed the most to the classification households based on the food poverty line in West Java Province.