Implementasi CRISP-DM Pada Analisis Pembangunan Pendidikan Prasekolah Menurut Kabupaten/Kota di Indonesia

Abstract

Preschool education through Kindergarten (TK) plays a crucial role in child development in Indonesia, yet unequal access remains a significant issue. This study evaluates the need for preschool infrastructure development using the K-Means clustering algorithm implemented through RapidMiner. Regional clustering is based on the number of students, number of TK schools, Human Development Index (HDI), poverty rate, population size, and unemployment rate. The CRISP-DM methodology is applied, involving stages of understanding, preparation, modeling, evaluation, and deployment. Data from the Central Bureau of Statistics (BPS) and the Ministry of Education's Dapodik system are utilized, incorporating Z-transformation normalization and data cleansing. The clustering results reveal three main clusters with the lowest Davies-Bouldin Index (DBI) at K=3, scoring 0.205. With a total of 514 districts/cities in Indonesia, the results of the needs of each cluster were obtained, namely Cluster 0 consisting of 402 districts/cities requiring increased participation, Cluster 1 covering 49 districts/cities requiring educational facilities, Cluster 2 covering 63 districts/cities requiring the construction of new schools. This study provides valuable insights into addressing disparities in preschool education access and offers guidance for better resource allocation and policy decisions aimed at improving early childhood education infrastructure.