Utilizing Naive Bayes for Multidimensional Poverty Classification and Identifying Key Poverty Determinants

Poverty assessment based solely on income is not enough measure of overall well-being, necessitating a multidimensional approach considering factors such as place of living, education, and employment quality. The Philippines faces challenges in accurately identifying poor households and individuals, leading to a significant mismatch between targeting criteria and actual beneficiaries in poverty alleviation programs. This study proposes an innovative approach utilizing machine learning, specifically a Naive Bayes classifier, to improve the quality of poverty assessment. Leveraging socio-economic indicators, the research aims to inform policy decisions and optimize resource allocation for poverty alleviation programs. The Naive Bayes classifier is compared with other machine learning models, demonstrating superior performance in predicting poverty status. The study's results reveal key indicators influencing poverty, including country, living area (urban or rural), and education level. The Naive Bayes classifier achieves a 69% balanced accuracy on a test set of unseen data. These findings contribute to the ongoing discourse on effective poverty alleviation strategies, emphasizing the potential of machine learning in addressing complex socioeconomic challenges.