Clustering of High-Achieving Students Based on Scores at Junior High School Level Using K-Means Algorithm

Abstract

Education plays a crucial role in shaping quality human resources, and student achievement evaluation at the junior high school level is essential for supporting academic guidance, learning programs, and recognition of outstanding students. However, the increasing number of students often makes the process of identifying and categorizing achievement more complex. This study aims to develop a student clustering model at SMP Budi Utomo Binjai using the K-Means algorithm as part of a data mining approach. The input data consisted of 638 student records covering three main variables: average score, counseling score, and extracurricular score. Data were preprocessed and transformed before being processed using MATLAB R2014a, which provides a kmeans() function to automatically group the data into clusters. Several clustering trials were conducted with three to six clusters to evaluate the grouping performance. The results showed that students could be grouped into categories of high, medium, and low achievement, with each cluster having different characteristics of average, counseling, and extracurricular scores. Variance analysis indicated that clusters with smaller variance values represented more compact and homogeneous groupings, while clusters with higher variance values were more heterogeneous. The findings demonstrate that the K-Means algorithm is effective in grouping student performance data objectively, providing useful insights for teachers and school administrators to design more targeted learning strategies, academic interventions, and recognition systems. This research highlights the potential of data mining techniques to support decision-making processes in the education sector.