Collaborative Filtering Recommendation for MOOC Application

With the fast development of MOOC in recent years, one MOOC platform has more than millions of learners and thousands of courses. Personality course recommendation is necessary to help people discover potentially interested courses by analyzing user behaviors. Collaborative Filtering (CF) has been shown to be effective on recommending items according to users in same interests on same items. In this paper, considering the effectiveness and efficiency of CF, we proposed a method called Multi-Layer Bucketing Recommendation (MLBR) to recommend courses on MOOC. MLBR changes learner vectors into same length dimension and scatters them into buckets which contain similar learners with more courses in common. At the same time, MLBR reduces the time cost of online, offline and update computation in CF recommendation. Furthermore, we extend MLBR with map-reduce technique to improve the efficiency. Extensive experiments on real-world MOOC datasets demonstrate the effectiveness and efficiency of the proposed model. (C) 2017 Wiley Periodicals, Inc.