CLASSIFYING STUDENT COLLABORATION PATTERNS USING K-MEANS CLUSTERING AND DECISION TREE ANALYSIS IN AN ONLINE LEARNING ENVIRONMENT

Abstract

Collaborative learning has become essential for teaching, learning patterns, and procedures with the rapid growth of E-Learning. Vigorous research over time revealed that student involvement in classroom discussions is significant and collaborative in acquiring knowledge. Collaborative learning is a method where a group of students works together to achieve a goal. TRAC is one of the online collaborative learning tools used by student teams. It is an open-access, professional software development tracking system. It supports collaboration by integrating three tools: a group wiki, a Ticketing system, and Subversion control. Data is collected from the student’s use of the TRAC tool by considering the wiki’s events, wiki pages edited, ticketing events, and subversion commits, all of which are traces of the student's actions. This paper aims to cluster students based on their performance in group coordination. The students are grouped to differentiate the better groups from the inferior groups. In this paper, K-Means clustering, an unsupervised learning algorithm, is used to group students and improve group performance. After clustering, evaluate the decision tree's performance on the train set with class labels and the test set without labels. It is a tree-like diagram that represents the outcomes in the leaf nodes. The performance is evaluated using accuracy, precision, recall, and F1-score, and compared against seven baseline classifiers. Scalability analysis confirms the approach generalizes effectively from 60 to 500 students. A pattern-recognition neural network is also employed to cross-validate the classification results. The results clearly show that the students belong to their respective groups.