Perbandingan Algoritma Clustering Untuk Optimalisasi Penerima Bantuan Kartu Indonesia Pintar (KIP-K) Di Wilayah Indonesia

Abstract

Poverty and limited access to education remain major issues in Indonesia, particularly in determining the recipients of educational assistance such as the Kartu Indonesia Pintar Kuliah (KIP-K). This study aims to optimize the determination of KIP-K recipients by using regional clustering based on socio-economic indicators, such as poverty levels and school dropout rates. The primary objective of this research is to identify areas that require more attention in the distribution of KIP-K assistance, ensuring it is more accurately targeted. Three clustering algorithms used in this study are K-Means, DBSCAN, and Agglomerative Clustering. The analysis results show that Agglomerative Clustering produces more distinct and cohesive clusters, with the highest Silhouette score of 0.6144, indicating clearer regional divisions. Meanwhile, K-Means, with a Silhouette score of 0.6128, also delivers satisfactory results, although not as optimal as Agglomerative Clustering. DBSCAN shows limitations in grouping data, as most regions are identified as noise. As an implementation of this research, an interactive Streamlit dashboard has been developed to visualize the clustering results and facilitate decision-making in the distribution of KIP-K assistance. This dashboard serves as an effective tool to enhance the transparency and efficiency of KIP-K distribution, supporting more precise decision-making.