An Approach for Early Prediction of Academic Procrastination in e-Learning Environment

Abstract

Ubiquitous learning is a new educational paradigm partly created by the affordance of digital media. This trend has continued to expand over time. The emergence of ubiquitous computing has created unique conditions for people working as education professionals and learning as students. Procrastination is one of the characteristics that has been seen in students that forces them to set back and sit back without achieving their goals. It has been estimated that almost 70% of college students or even school students engage in frequent academic procrastination and purposive delays in the beginning or completing tasks. Throughout this study, we concentrated on different predictive measures that can be used to identify procrastination behaviour among students. These measures include the usage of ensemble classification models such as Logistic Regression, Stochastic Gradient Descent, K-Nearest Neighbours, Decision Tree and Random Forest. Of these, the random forest model achieved the best predictive outcome with an accuracy of almost 85%. Moreover, earlier prediction of such procrastination behaviours would assist tutors in classifying students before completing any task or homework which is a useful path for developing sustainability in the learning process. A strength of this study is that the parameters discussed can be well defined in both virtual and traditional learning environments. However, the parameters defining students’ cognitive or emotional states were not explored in this study.