A Study Of Collaborative Filtering Recommendation Algorithm

With the development of Internet,network information resources have increased exponentially.Faced with huge amounts of information resources,it is difficult for users to find the resources which they need.The users are faced with the problem of information overload.In order to help user find more easily the information resources they need,the recommendation algorithm arises at the historic moment and is widely used in various recommender systems.Collaborative filtering recommendation algorithm is the most successful recommendation algorithm at present.It is mainly based on the historical behavior information of the target user in the recommended system,and calculates the similarity between the other users and the target users.The user sets which are most similar to the target users will determine the information resources recommended to the target users.Similarity calculation are dependent on historical behavior information,which is divided into user-item rating information and user-trust rating information.To address the problem of sparseness of user rating data,this paper proposes a recommendation algorithm based on user ratings and item attribute preferences.This paper improves two aspects of the traditional recommendation algorithm.On the one hand,this paper constructs the similarity based on item attributes from the respect of the users' preferences for project attributes.This paper constructs the similarity based on item rating from respects of the difference of item rating ? rating habits and reliability.This paper then combines the two similarities.On the other hand,this paper changes the search space of the recommendation algorithm.Using the clustering algorithm based on the preferences for project attributes and the clustering algorithm based on the item rating can obtain two similar clusters which are most similar to the target user.This paper combines the two clusters as the search space for the nearest neighbor set.Experiments show that the algorithm can alleviate the problem of the sparsity of data to some extent,and has higher recommendation accuracy than the traditional collaborative filtering recommendation algorithm.To address the problem of sparseness of user trust relationship,this paper proposes a method to extend trust relationship.This method fully considers the influence of various factors on trust relationship and extends the initial trust relationship.Then,this paper combines trust relationship and similarity and propose collaborative filtering recommendation algorithm based on trust similarity.Experiments show that compared with the traditional recommendation algorithm based on trust,the algorithm has higher recommendation accuracy and higher coverage.