Multiclass student engagement level prediction using belief-rule based labelling

Abstract

Educational Data Mining (EDM) is the process of understanding the student learning method, studying educational questions, and enhancing the teaching and learning (T&L) process. Online learning (OL) delivery will only be successful when the student successfully receives it. Therefore, student engagement with OL can lead to student success and enhance student satisfaction, motivation, and performance. Student engagement is a multidimensional construct made up of behavioural, cognitive, emotional, and social engagement. Therefore, the belief rule based (BRB) is proposed to label student engagement into three different levels in this research. Then, five classification algorithms which include Decision Tree (DT), Random Forest (RF), Naïve Bayes (NB), Support Vector Machine (SVM), and Logistic Regression (LR). The proposed BRB successfully removes the noise sample and improves the quality of the dataset. Among all the five classification algorithms, SVM performs the best for the prediction of student engagement levels giving 90% accuracy.