Feature Engineering for Predicting Student Dropout in Massive Open Online Courses

Abstract

Massive open online courses (MOOCs) were invented to impart education to people who could not afford or get access to traditional education options. In recent times there is an unprecedented shift to online platforms for teaching-learning around the world. Dropout prediction or identifying students at risk of dropping out of a course, therefore becomes an important problem to study due to the high attrition rate commonly found on many MOOC platforms. Proper feature engineering plays an important role in getting desired results in such predictions. The present work thus focuses on providing valuable features for improving the prediction results. These predictions are then performed and analyzed by Machine Learning algorithms. The Random Forest Classifier gave 86.14% accuracy and outperformed Logistic Regression, Decision Tree Classifier, Gradient Boosting Classifier and Xgb Classifier applied in the present work.