Explainable Machine Learning Approach to Classifying Poverty Levels Across Indonesia

Abstract

Poverty in Indonesia is a complex and multidimensional socio-economic problem, influenced by factors such as education, expenditure, and access to basic resources. This study aims to comprehensively classify the poverty level in Indonesia by utilizing a Machine Learning approach based on the Random Forest algorithm combined with the SHAP (SHapley Additive exPlanations) interpretability method. Research data obtained from the Central Statistics Agency (BPS) for the 2023–2024 period, includes 514 provincial and regency/city records with 12 socio-economic variables such as the percentage of the poor population, average length of schooling, per capita expenditure, Human Development Index (HDI), access to clean water and sanitation, and open unemployment rate. The data is divided into 70% for training and 30% for testing. The results of the evaluation showed that the Random Forest Classifier model achieved an accuracy of 96.13%, precision of 96.29%, recall of 96.13%, and an F1-score of 96.17%, showing excellent performance in classifying poverty into five classes. The interpretability analysis using SHAP identified that the most influential features on the poverty rate were the percentage of the poor population (P0), per capita expenditure, and average length of schooling. Increased education and per capita expenditure reduce the risk of poverty, while high P0 values and unemployment increase the likelihood of a region being classified as poor. The results of this study not only produce accurate prediction models, but also provide transparent and actionable insights for policymakers in designing data-based, sustainable, and targeted poverty alleviation strategies in Indonesia.