Bridging the gap: regional poverty prediction and policy insights using machine learning

Poverty remains a significant socio-economic challenge in Malaysia, with regional disparities affecting the effectiveness of centralized aid programs. Many interventions lack precision, often due to the limited use of predictive tools and insufficient data-driven targeting. This project introduces a machine learning-based system for classifying poverty levels (low, moderate, high) using multi-domain socio-economic indicators, including income, employment, education, and inflation. Publicly available data from the Department of Statistics Malaysia (DOSM) from 2019 to 2022 was used to train and evaluate five classification models. Among them, the XGBoost model demonstrated the highest performance, achieving an accuracy of 77.2% and an AUC score of 0.93. Key features influencing poverty were selected based on the model that performed best. The solution is deployed as an interactive web application using Streamlit. The dashboard provides district-level insights into poverty, visualisations of poverty distribution, and a feature importance analysis. It also presents tailored policy recommendations based on dominant poverty factors in each region. This tool empowers decision-makers to implement more informed and evidence-based poverty interventions. By bridging artificial intelligence with social policy, the project contributes to more effective resource allocation and equitable development strategies across Malaysian states.