Poverty risk prediction based on socioeconomic factors using machine learning approach

Poverty remains a persistent socio-economic challenge in Malaysia, affecting access to basic needs such as healthcare, education, and employment. Traditional measurement methods like the Poverty Line Income (PLI) often oversimplify poverty classification by relying solely on income, ignoring multidimensional factors that influence well-being. This study aims to develop a predictive model for assessing poverty risk using supervised machine learning techniques based on a wider range of socioeconomic factors. A secondary dataset involving 635 households from eight districts in Terengganu was analyzed. Attributes such as age, income, education, occupation, health, gender, marital status, and savings were considered. Data preprocessing, including cleaning and transformation, was conducted, followed by feature selection using the Information Gain method. Four machine learning classifiers, such as Logistic Regression, Random Forest, J48, and Logit Boost, were evaluated using 10-fold cross-validation and a 70:30 training-testing split. Logistic Regression achieved the best overall performance with 99.06% accuracy in cross-validation and 98.42% in data splitting, along with superior precision, recall, and F1-score values. Age was identified as the most significant factor of poverty risk, followed by income and occupation. These results demonstrate that Logistic Regression is a reliable, interpretable, and stable model for classifying poverty risk using structured socioeconomic data. Although the study is geographically limited, its findings support the potential of machine learning as a data-driven tool in social policy development. This approach may enhance poverty targeting strategies and contribute to achieving Malaysia’s Sustainable Development Goals (SDG) 1: No Poverty and SDG 2: Zero Hunger.