Improving Prediction of MOOCs Student Dropout Using a Feature Engineering Approach

Recently, the realm of education has been revolutionized by open massive online courses (MOOC). They have gained more importance and interest and greatly evolved as they provide a way of learning chiefly free online users around the world by millions of participants. Although MOOCs boast several characteristics and benefits, they have a major pitfall associated with high dropout rate. The analysis of MOOC data gives useful tools of shedding light on the characteristics that can facilitate the understanding of the behavior of the learners and accompany them in order to make their learning successful. In this paper, we explore the application of different data science techniques, including feature engineering and predictive modeling, to identify a student who is likely to dropout, utilizing the data from the KDD 15 with several supervised classification models. Two types of experiments were conducted. In the first set of experiments, all the features are used, and passed to the ML, while in the second set of experiments, only high ranked features are used. Our experiment gives the best accuracy in the dropout prediction task with GBDT model with high ranked features.