Research And Application Of Collaborative Filtering Algorithm Based On Coupling Similarity

With the rapid development of Internet and e-commerce,recommendation systems gain more and more attention and experts and scholars gradually set their sights on how to improve the performance of the recommendation system-s.In the process, a variety of excellent algorithms and models have been proposed, tested and put into application. Since Yehuda applied matrix factorization for recommendation system in Netflix Prize for the first time, there have been lots of improvement and perfection in recent years, which some algorithms for improved basis model is by using the additional information, such as bias, implicit feed-back and so on. We find that the traditional matrix factorization models often ignore consideration of the user’s property information or item’property infor-mation.However,we often say these properties contains a lot of useful information which helps personal recommendation a lot.Now our purpose in this paper is to taking the attribute of users and items into consideration to improve the rec-ommendation accuracy in the collaborative filtering model(CF model).Generally two types of recommender systems have been investigated:Memory-based and Model-based recommenders.Matrix factorization is one of Model-based recom-menders.We also see that some data is numerical, other is non-property in the at-tribute information of users and items,for which the concept of coupled similarity is introduced.Then we compute the attribute information of users and items for the establishment of similarity model.Further,similar information will be integrat-ed into the CF model.At last,we test the effectiveness of the algorithm proposed in this paper in real data sets and compare the result with other recommendation technology, what’s more,we also get high recommendation accuracy in the "cold start" and "sparsity".In this paper, there are as follows:The first chapter is the introduction, introducing the research background and the main content; The second chapter is about relevant knowledge, introduces the collaborative filtering algorithm; the third chapter is about coupled similarity concepts,the improved Memory-based model proposed in this paper and the experiment and discussion of the algorithm proposed in this chapter;the fourth chapter is about framework of the improved Model-based model and the the experiment and discussion of the algorithm pro-posed in a real data sets, and other recommended technical comparisons. And there are the summary and outlook at last.