Predicting MOOC Dropout Using Ensemble Models Based on RNN and GRU

Nowadays, dropout prediction in Massive Open Online Courses (MOOCs) has become a popular field in educational data mining. In order to predict early dropouts with high accuracy, this research introduces a model for data pre-processing. In addition, since our data consist of time series, Recurrent Neural Networks (RNNs) and Gated Recurrent Units (GRUs) are used. The study proposes six ensemble models named Adaboost RNN, Voting RNN, Bagging RNN, Adaboost GRU, Voting GRU, and Bagging GRU to improve the accuracy of dropout prediction. The models were trained and evaluated using KDD Cup 2015 data set. Results show that the ensemble models outperform both RNN and GRU models individually, with Bagging GRU achieving an accuracy of 91.73% and improving precision, recall, and F-score as well. The findings suggest that deep learning techniques such as RNN and GRU can be effective tools for predicting MOOC dropout rates, and that ensemble models can further enhance their predictive power.