K-Means Clustering for Prospective Student Interest Segmentation Based on Study Program Preferences

Abstract

Higher education institutions in the Labuhanbatu Raya region face challenges in understanding the interests of prospective students. Data collected from 1,722 prospective students at 29 high schools has been stored but has not yet been optimally utilized to support student recruitment and admissions strategies. This study aims to apply the K-Means algorithm to cluster prospective students’ interests based on their preferred degree programs. The research methodology included data collection, feature selection, data cleaning, transforming categorical data to numerical using label encoding, determining the optimal number of clusters using the Elbow method, clustering with the K-Means algorithm, evaluating cluster quality using the Silhouette Score, and visualizing and interpreting the clustering results. The findings indicate that the optimal number of clusters is K = 3, with a Silhouette Score of 0.3695, indicating fairly good clustering quality. The cluster distribution shows that Cluster 0 consists of 19 schools (65.52%) with relatively homogeneous interest patterns, Cluster 1 consists of 9 schools (31.03%) with more diverse interest patterns, and Cluster 2 consists of only 1 school (3.45%) with a very distinct interest profile. The K-Means algorithm proved effective in clustering schools based on prospective students’ program preferences.