Learning Resource Recommendation System Based On E-learning User Behavior

Nowadays,the way people learn is no longer confined to the classroom and books.With the rapid development of mobile Internet and communication technology,learning method extends to varieties of E-learning platform,which supports online learning via mobile phones or computers.In order to meet the changing needs of people,the recommendation technology has been gradually applied to the learning platforms.However,in the process of recommending learning resources for learners,most of the learning platforms are keen to recommend the popular learning resources to learners,ignoring the learner's individual learning needs.Thus,it is an important research topic that how to provide the learners with better learning resources in E-learning.In this thesis,collaborative filtering recommendation technology is chosen as the basic algorithm to recommend learning resources.On the one hand,with the increase of the huge learning resource and user cardinal number in the learning platform,the score cardinal number of learning resources is getting smaller and smaller,which leads to the problem of data sparsity in the recommendation algorithm.On the other hand,the behavior of users in the system is a huge wealth for the learning analysis,however,these behaviors are often scattered in every corner of the system and cannot take them full effect.In view of the above problems,we study the algorithm level and data acquisition level respectively.The main research work is as follows: 1.Aiming at the problem of data sparsity in E-learning platform,collaborative filtering is improved from the similarity calculation method in this thesis.In the measure of the similarity between users with a common score,we introduce the concept of similarity factor,not only consider the similarity between users,but also take into account the number of joint scoring resources.In the measure of the similarity between users with no common score,we propose the similarity transitivity scheme,and use the similarity factor after transmission to measure the similarity between users.We validate the algorithm by using the data of "YiBi" platform of Chongqing digital media education project.The results show that the algorithm has achieved good results in prediction accuracy.2.In order to enrich the data source,using the xAPI data normalization to track and collect the learning behaviors of learners in the learning platform,furthermore,on the basis of the improved algorithm in this thesis,we design and implement a personalized recommendation system based on xAPI model.