Penerapan Algoritma K-Means Untuk Klusterisasi Penerima Dana Bantuan Pendidikan KJP Plus Pada DKI Jakarta

Abstract

This study aims to develop a clustering model of Jakarta Smart Card (KJP) Plus Phase 2 beneficiaries for the 2025 fiscal year based on transportation asset ownership and school status, in order to improve the targeting accuracy of educational assistance as highlighted by the audit findings of the Indonesian Supreme Audit Agency (BPK). A database of 707,513 beneficiaries was merged with 8,958,396 motor-vehicle tax records through matching of the guardian's national identity number, yielding 398,590 beneficiaries identified as vehicle owners. Clustering was performed using the K-Means algorithm in the Orange software with two attributes, namely vehicle weight and school status, the latter first expanded into binary columns through one-hot encoding. The optimal number of clusters was set at two (k = 2) based on the highest Silhouette coefficient of 0.836, confirmed by a Davies-Bouldin Index of 0.41. The results reveal two groups distinguished by vehicle type: Cluster 1 (393,909 beneficiaries; 98.8%) consists of light-vehicle owners dominated by motorcycles, whereas Cluster 2 (4,681 beneficiaries; 1.2%) consists of owners of four-wheeled and commercial vehicles; school status does not differentiate the clusters. The study concludes that the small group of heavy-vehicle owners represents the strongest indication of ineligibility and warrants priority for verification.

GS Cited Analysis