Implementation of K-Means Clustering for Optimization of Student Grouping Based on Index of Learning Styles in Programming Classes

Abstract

This study aims to group students into study groups (classes) based on learning styles utilising K-Means Clustering technique  and Sum of Squared Error for cluster assessment. This study used type of learning style developed by Felder and Silverman, which includes four dimensions: (1) the learning process; (2) perception of learning; (3) information input; and (4) understanding of information. This study subjects were Universitas Sebelas Maret's students majoring in informatics Education consisting of 58 respondents. The results showed that the K-Means clustering approach with cluster evaluation using Sum of Squared Error produced the best clustering when the number of clusters was k=2. The cluster analysis showed that each class has different learning styles and characteristics. The first cluster (group) consists of 26 respondents and has the features of an active learning style in the learning process dimension, sensing in the learning perception dimension, visual in the information input dimension, and a balance between global and sequential in the information understanding dimension. Meanwhile, the second cluster (group)  consists of 32 respondents and has a reflective tendency in the learning process dimension, sensing in the learning perception dimension, visual in the information input dimension, and global in the information understanding dimension.