Research And Application On Trust-Based Collaborative Filtering Recommendation

The problem of information overload is an inevitable problem in the process of rapid development of the Internet and the Internet service application,which makes it more difficult for users to collect and obtain information quickly and effectively.Personalized recommendation system is an effective means to solve this problem,it analyses the user's interest or preferences and the link between the goods through the information of the user or the commodity,and then realizes the user's personalized recommendation service.As one of the more mature and more widely used recommendation technologies,collaborative filtering recommendation technology has the advantages of low data dependence,high accuracy and easy calculation.However,on the one hand,it only considers user's interest or preferences one single factor,on the other hand,it faces the bottleneck of data sparsity,cold start,scalability,security,and so on,which greatly affect the recommendation of collaborative filtering.The development of online social networks has brought new opportunities for the study of the recommendation system.As the core and key of online social networks,trust has been introduced into the construction of recommendation system more and more.Compared to the traditional personalized recommendation,due to the comprehensive consideration of user's interest or preferences and social relations information two dimensional factors,recommendation based on trust have better results in accuracy,reliability,security and other indicators and further ease the collaborative filtering recommendation technology and other traditional recommendation technology which are faced with data sparsity,cold start and robustness issues,etc.In this paper,we focus on how to use both trust in online social networks and collaborative filtering recommendation technology to improve the performance of the recommendation system.Firstly,By reorganizing and analyzing the domestic and foreign studies and theories of trust in online social network and collaborative filtering recommendation,this paper systematically reviewed the theories related to the recommendation based on trust,recommendation based on collaborative filtering and discussed the relationship between them.Secondly,an improvement of trust propagation model is put forward,which can realize the extension of the trust network and solve the problem of sparsity of trust,and the trust degree is modified to better reflect the characteristics of trust between users.Thirdly,based on the improved trust model,this paper proposes an improved trust based collaborative filtering recommendation model(based on the Trust Walker recommendation model).By making full use of the trust relationship and interest to solve the problem of data sparsity,cold start problem and attack of the traditional personalizedrecommendation system,and improves the accuracy,coverage and credibility.Then,based on a certain rule of data selection and preprocessing,this paper select a part of the data set of Epinion.com for the experimental analysis of the proposed model.Experimental results show that,by using the trust information and trust model,we can find more potential users,improve the coverage and credibility of the recommendation,alleviate the data sparsity and other issues,and compared to the Trust Walker recommendation model,the improved model in this paper has better effect on MAE index and PS index.Then,design a simple mobile terminal based movie recommendation system and build a basic framework for the practical application of trust based collaborative filtering recommendation.In the end,the research content of this paper is summarized,the deficiency of the model is analyzed,and the next step is discussed.