Implementation of K-Means Clustering for Social Assistance Recipients with Silhouette Score Evaluation

Abstract

The distribution of direct social assistance continues to face several challenges, particularly regarding inaccurate targeting and unequal allocation. One of the main causes of this issue is the lack of transparency in the distribution process, where assistance is often granted to individuals with familial ties to local committee members or even government officials. As a result, the groups most in need frequently do not receive the aid they deserve. This condition is also evident in Tanjung Village, Bajuin Subdistrict, Tanah Laut Regency. The manual process of grouping prospective aid recipients contributes to inaccuracies in targeting, which in turn leads to public dissatisfaction. To address this issue, this study applies the K-Means Clustering method to group potential social assistance recipients using data from 150 individuals and three main attributes: age, occupation, and income. The method clusters the data based on the similarity of characteristics, thus supporting a more equitable and efficient identification process. The evaluation is conducted using the Silhouette Coefficient to assess the quality of clustering. The results indicate that the highest Silhouette Score is achieved at k=2k = 2k=2, with a value of 0.8278, suggesting that dividing the data into two clusters provides the most optimal configuration. The Silhouette Score tends to decrease as the number of clusters increases, confirming that adding more clusters does not necessarily improve the quality of separation.