Application of Machine Learning for Clustering Academic Achievement Levels of Elementary School Students

Abstract

Academic achievement clustering provides an opportunity to identify student performance
patterns and support data-driven educational evaluation. This study aims to apply the K-Means
Clustering algorithm to group academic achievement levels among students at Al Muttaqien
Private Elementary School. The research employed a quantitative approach using academic
records from Grades 1 to 6 covering two semesters. Seven subject scores were selected as
numerical features, including Islamic Religious Education, Pancasila, Indonesian Language,
English, Mathematics, Physical Education, and Arts and Culture. The dataset was processed
through data cleaning, feature selection, and standardization before the clustering algorithm
was implemented. A total of 311 student-semester records were included in the analysis, and
a three-cluster solution was examined to identify relatively lower, moderate, and higher
academic achievement profiles. The results indicated average academic scores of 76.51,
83.32, and 89.29 for the respective clusters, representing 26.4%, 37.9%, and 35.7% of the
analyzed records. Cluster evaluation using the Silhouette Score showed that the three-cluster
solution achieved a score of 0.387, while the two-cluster solution achieved a higher score of
0.484. These findings indicate that K-Means can provide an interpretable grouping of
academic performance, although the selection of the optimal number of clusters requires
consideration of both statistical validation and educational objectives. The resulting profiles
may support teachers in identifying differences in student achievement and planning
differentiated learning strategies. This study demonstrates the potential of unsupervised
machine learning as a supporting approach for academic data analysis in elementary
education.