Ability Similarity Measure of MOOC Learners Based on Video Learning Data

Similarity measure of MOOC learners is a hot topic in the current research of educational data mining, and it is also the basis of learners clustering and grouping. In the online education environment based on MOOC, learning MOOC videos is one of the most basic behaviors of learners. The degree and ability of learners to master the videos’ content can be implicitly obtained through their video learning behavior, thus providing a basis for the measure of learners’ ability similarity. Most existing researches on the similarity of learners focus on the similarity of learners’ interests or behavior patterns, and the similarity measure of ability is ignored. Meanwhile, most existing works only use video related data as a dimension of learners’ similarity measure, and there are still shortcomings in judging the ability similarity of learners. This paper proposes an approach to measure learners’ ability similarity based on MOOC video learning data. Based on the videos and their learning times of learners, a bipartite graph model is constructed, and the ability similarity between all learners is measured iteratively through SimRank++ algorithm. The experiments based on the real data set show that the proposed approach has better accuracy than the cosine similarity that is widely used in related works, and the NDCG value is increased by 34% on average.