Research On Collaborative Filtering Recommendation Algorithm Based On User Interest Similarity And Trust Mechanism

With the fast development of the Internet for the last few years,especially in the field of Internet economy,there are more and more ways and means for users to obtain information in the internet.With it,it is easy to cause the information overload phenomenon caused by the explosive increase in the amount of information.In order to convenient for the user in the huge amounts of information quickly find useful information,which can meet their requirement,solve the problem of too much data and cause of overload,recommendation system certain trendency,quickly find useful information that meets their own requirement in the massive information,and solve the overload phenomenon,the recommendation system emerge and get fast development.In spite of the recommendation algorithm is regarded as is the most widely used,one of the most popular algorithms at present,but there will still exist two problems for researchers to solve.Based on the core of the recommendation system for the gripper,and the multi-dimensional similar degrees of improvement in this paper,research on collaborative filtering recommendation algorithm based on user interest similarity and trust mechanism is proposed.(1)Aiming at the shortcomings of traditional collaborative filtering algorithms,an optimization idea to improve the user interest form the user's preference to the item attribute characteristics and the user attribute characteristics form which itself comes are made.Where in the user attribute similarity modification,the angle tends to extract from an individual user's age,gender,occupation,and wherein three attributes,wherein each attribute and to impart a certain weight,considering the similarity between the user characteristic properties;At the same time,by using feature extraction technique of TF-IDF,which based on content recommendation,the user's feature words for item attributes are extractd.And then calculates the weight of each attribute to reflect the user's preference for the item attribute.Finally,the user's preference to the item attribute characteristics and the user attribute characteristics form which itself comes are fused to obtain a better optimized user interest similarity,and achieve more accurate recommendation.(2)Taking into account only rely on the user rating data,so that a single similarity calculation result,and when the user does not produce scores,recommendation results can not be guaranteed.Thus,in order to abtain the optimized similarity,the paper need to introduce the trust mechanism between users,the algorithm uses the implicit factor and display factor get combination,and propose effective trust evaluation reflects direct trust between the users.On the other hand,by combing the direct trust and indirect trust,the tust relationgship between users has been expanded,obtains a comprehensive trust between users,and integrates multiple linear similarities to obtain an improved algorithm that makes the recommendation result more credibility.To sum up,papers to solve user rating sparsity and cold-start problem,the paper will comprehensively consider the user's preference to the item attribute characteristics and the user multiple attribute characteristics form which itself comes,which can estabish a user interest similarity based on multi-layer similarity fusion.Then,by combing the direct trust and indirect trust can improve the mechanism of trust between users,and the tust relationgship between users has been expanded in the end.Combing several similarity calculations of the above we talked in this paper,research on collaborative filtering recommendation algorithm based on user interest similarity and trust mechanism is proposed.,and put this algorithm into the experiment of Movielens movie data to evdient the algorithm that we put forward can improve the accuracy in some extent.