Identifikasi Pola Prilaku Belajar Mahasiswa Pada Platform Learning Management System Dengan Algoritma K-Means

Abstract

This study aims to explain students' behavior in accessing the kalam.umi.ac.id learning platform and its impact on student academic performance. It analyzes data collected from students' access to the platform during 12 sessions in one semester for the Islamic communication subject taught by Professor Surani. The data includes students' access times, access frequency, and academic performance such as assignment grades, midterm exams, and final exams. Through data processing methods, correlation analysis, and cluster optimization, the study found a positive relationship between access times, midterm exams, final exams, and assignments with students' final grades. The higher these variables, the higher the students' final grades. However, this relationship is not always consistent across different variables. In this study, the Elbow method was used to determine the optimal number of clusters by identifying the point where the variance reduction becomes less significant. Additionally, the Sum of Square Error (SSE) was analyzed to understand the sharp change followed by a gradual decrease in the value of K until stability is reached. Clustering results using the K-Means algorithm showed the presence of three student clusters based on their learning behavior. Cluster 0 is the largest, consisting of 176 students. Cluster 1 has 57 students, and cluster 2 is the smallest with 17 students. These clusters provide insights into varying student learning patterns, including differences in final grades and access frequency. These findings can be used as a basis for developing more effective and personalized learning strategies for students.