An Early Prediction of Dropouts for At-risk Scholars in MOOCs using Deep Learning

Abstract

In 2012, the world has observed a new approach to educational systems named ”Massive Open Online Courses” which leads to a greater impact on the entire world as it transformed our traditional educational approaches. MOOCs have rapidly moved into a place of prominence within the media, in scholarly publications, and within the mind of the general public Universities. According to recent studies, more than thirty MOOC platforms are providing ethnically diverse courses to scholars across the boundaries however it was also observed that they can hardly follow the subjects until the end of the course. Such conclusions pointed out that in the future it may restrict the flourishing of such platforms and these high dropout rates may diminish the development of MOOCs. The research work focuses on developing an early prediction of a dropout system for at-risk scholars in MOOCs using the deep learning algorithm and simulating a dropout prediction model to construct a ranking system for scholars using their dropout probability for every week. Using this probability value the system could be able to predict the scholar’s attitude toward the subject and chances of dropout; thereby this system would help the instructors to guide them at an early stage. The datasets of MITx and HarvardX MOOC courses were used to predict scholar dropouts by analyzing their learning patterns. The Deep Neural Network model could make predictions better than the existing technologies through hyperparameter tuning and optimization. Inducing the benefits of deep learning methods helps to construct an effective dropout prediction model and this new versatile technique helps the tutors to prioritize intervention for those dropouts and can be considered an effective solution to the early dropout students in MOOCs compared to the existing works.