Recommendation for MOOC with Learner Neighbors and Learning Series

MOOCs (Massive Open Online Courses) have become increasingly popular in recent years. Learning item recommendation in MOOCs is of great significance, which can help learners select the best contents from the huge overloaded information. However, the recommendation is challenging, since there’s a high percentage of drop-out due to low satisfaction. Not like traditional recommendation task, learner satisfaction plays an important role in course engagement. The lower the satisfaction is, the higher possibility the learner would drop out the course. Aiming at this, we propose a new recommendation model-Recommendation with learner neighbors and learning series, called RLNLS. It takes achievement motivation on satisfaction into account by exploiting and predicting learning features. A new feature model aiming at satisfaction is proposed according to Expectancy-value Theory. More specifically, knowledge distance is presented to prediction of learning features with learner neighbors and learning series. Hawkes process is modified and utilized for learning intensity prediction. The experimental results on real-world data show the effectiveness of the proposed model in recommending courses and reducing drop-out rate by a large margin.