Research On Collaborative Filtering Recommendation Algorithm Based On Resource Characteristic

The purpose of constructing education resource library is to provide education services which take learners as the main body, so it should consider the individual differences of different learners, and then recommend resources to different users with their individual characteristics. The main target of this thesis is to establish a high-quality recommendation engine of basic education resources to meet the personalized needs of teachers. This thesis mainly does a research on the collaborative filtering recommendation algorithm to make it meet the individual requirements of basic education resource library.Firstly, this thesis briefly introduces the research background and significance of the task, as well as the components and current research status of recommendation engine, and so on.Secondly, the thesis elaborates some of the most popular personalized recommendation algorithms, including their basic principles, processes of producing recommendations, advantages and disadvantages, and so on.Thirdly, aiming at shortages of previous collaborative filtering recommendation algorithms and combining the actual situation of basic education resource library, this paper puts forward a method of collaborative filtering recommendation based on resource characteristics. It turns user’s interest behaviors into user’s interestingness to keywords and turns the changes of user interest into the changes of the weights of user’s interested keywords. In this way, we can establish and update user interest model. At the end, it finds resources that target user may be interested in by finding out the similarities between different users or different resources.Finally, putting the cooperative filtering recommendation algorithm based on resource characteristics into the prototype system on the basis of basic education resource library to verify the validity of the algorithm. The test shows that this algorithm can not only track user’s interest drift, but also greatly alleviate data sparsity problem and new item problem and new user problem.