Implementation of a Machine Learning-Based MOOC Recommender System Using Learner Motivation Prediction
Sara Assami, Najima Daoudi, Rachida Ajhoun
Abstract
The phenomenon of high dropout rates has been the concern of MOOC providers and educators since the emergence of this disruptive technology in online learning. This led to the focus on learner motivation studies from different aspects: demotivation signs detection, learning path personalization, course recommendation, etc. Our paper aims to predict learner motivation for MOOCs to select the right MOOC for the right learner. So, we predict the motivation in an educational data mining approach by extracting and preprocessing learners' navigation traces on a MOOC platform and building a machine learning model that predicts accurately a given learner motivation for a MOOC. The comparison of the performance of four supervised learning algorithms resulted in the selection of the random forest classifier as a modeling technique for motivation prediction. Afterward, the Machine Learning-based recommendation function was tested for learners of the MOOC platform dataset to recommend the Top-10 MOOCs suitable for the target learner. Finally, further research on learner characteristics considered in recommender systems could enlarge the recommendation scope of MOOCs and maintain learner motivation.