Machine Learning Approaches to Digital Learning Performance Analysis

Abstract

Academic learning performance prediction is one of the concerns for the stakeholders of the educational system, namely administrators, teachers, students, parents, and others. As poor performance in learning may lead to drop out of a student, so it is vital to predicting the performance to identify the student at risk. By identifying the student at risk, corrective action can be taken in advance for the improvement of the performance. The purpose of this study is to identify the students who may have problems in the coming course sessions. These problems can lead to poor performance. In this study, we have performed comparative analysis for different machine learning algorithms named; Artificial Neural Network (ANN), Naïve Bayes (NB), Decision Tree (DT), Logistic regression (LR), and Support Vector Machine (SVM), on extracted features. The extracted features are average mouse clicks, total activities, average time, average idle time, average keystrokes, and total related activities in an exercise. The results exhibit that SVM is better to predict the performance as equated to other machine learning techniques, by the accuracy of 94.82 %. These findings can suggest measures to take action like additional help required in advance to a particular learner for the success at a higher level of Bloom’s Taxonomy.