Comparison Clustering Performance Based on Moodle Log Mining

Abstract

The use of digital data now saves a lot of information, but still raw and not in the form of knowledge that can be directly seen, therefore we need a data processing to get useful particular understanding from raw data. Data Mining has an important role to process and find useful information from data. This process is also called Knowledge Discovery. Educational Data Mining is a Knowledge Discovery process in the world of education using data mining techniques. One method used in data mining is clustering. By using clustering analysis techniques, data can be grouped into groups without the need for prior knowledge (previous knowledge), so that the data is grouped based on similarity of patterns. This paper compares the K-means, Hierarchical and Louvain clustering methods to see the most appropriate clustering technique in analyzing log activity data in Moodle Learning Management System (LMS). The results of clustering are measured using the Silhouette Coefficient, and then we compare the values and distribution between clusters. In conclusion, Hierarchical clustering produces the highest Silhouette Coefficient value and also this algorithm can detect outlier data as new cluster. Louvain clustering perform very well to find cluster groups in new dataset as the algorithm does not required the number of clusters to be specified before. Louvain clustering can divide more evenly and precision compare to K-means and hierarchical cluster techniques.