Different Machine Learning Models to Predict Dropouts in MOOCs

Massive Open Online Courses have emerged as an alternative to the traditional educational system because of the flexibility in timings and also it overcomes the economic and geographical barriers for the users. MOOCs also help learners from diverse background to communicate and exchange knowledge in MOOCs forums. The number of learners enrolling in such courses is very high. Despite the unrestricted accessibility, the completion rate is very low. Various factors affect the completion of the course by the students such as interest in the subject, the purpose of enrolling in the subject, whether the lecturer is able to convey his knowledge to the students or not. EDM (Educational Data Mining) and LA (Learning Analytics) are the fields in which data of students learning activity is analyzed to obtain certain vital information or can be used in prediction using EDM tools and techniques. Data analysis shows that there is a strong relationship between the number of events such as click event, video watched, forum post and the successful learner's outcome. Machine Learning algorithms are applied on the dataset from HarvardX and the result shows that Random Forest gives an optimum result with the highest performance.