Grouping the Eligibility Levels of Social Assistance Recipients Using the K-Means Clustering Algorithm

Abstract

Social assistance programmes (bansos) are government initiatives aimed at improving the welfare of low-income communities. However, the identification of potential beneficiaries often faces challenges, particularly regarding accuracy and targeting, due to a lack of systematic data analysis. Therefore, the use of data mining technology is essential to support more objective and measurable decision-making. This study utilises the K-Means Clustering algorithm to group the population by income level in order to identify patterns among social assistance recipients based on the proximity of their income attribute values. The research data comprised 1,666 entries of social assistance recipients in Malang City, with attributes including name, address and income. The analysis was conducted using the RapidMiner application through data pre-processing stages, including attribute selection and data normalisation using z-transformation to ensure all variables were on a uniform scale. Subsequently, the K-Means algorithm was applied to form three main clusters. The research findings indicate that cluster 0 represents the low-income group, cluster 1 the middle-income group, and cluster 2 the high-income group. These findings demonstrate that the K-Means Clustering algorithm is effective in mapping the economic conditions of the population and supports the distribution of social assistance in a manner that is more targeted and based on structured data analysis. It is hoped that this research will serve as a reference for government agencies in improving the transparency, accuracy and effectiveness of data-driven social assistance distribution policies that are fairer and more sustainable at the national level.