Identifying Student Learning Behavior Patterns Using K-Means Clustering

A Case Study At STMIK Mardira Indonesia

Abstract

Understanding student learning behavior is essential for developing effective instructional strategies and improving academic evaluation systems in higher education. This study aims to identify and characterize student learning behavior patterns using a clustering approach based on academic assessment data recorded in the Academic Information System (SIAKAD). To capture stable and long-term learning behavior tendencies, this research utilizes longitudinal academic records collected over eight consecutive semesters. The analyzed learning behavior attributes include assignment scores, quiz results, midterm examination scores, final examination scores, and attendance rates. The K-Means clustering algorithm was applied following data preprocessing and z-score standardization, while the optimal number of clusters was determined using the silhouette coefficient. The results reveal three distinct learning behavior patterns, namely students with low learning engagement, students with moderate engagement characterized by an exam-oriented learning strategy, and students with high and consistent learning engagement across learning activities.