A Service Recommendation System Research Based On Collaborative Filtering

The birth of the Internet has brought great convenience for the user,and it also produced a large number of data and information with the continuous application of the Internet at the same time.To service providers in the information in the ocean,extracting information from the user demand,efficient service ability requirements are put forward.With the continuous upgrade of these requirements,the user is in use process,the low efficiency problem,known as information overload.Therefore,researchers put forward the concept of the recommendation system which can help users to complete information filtering work.Recommendation system is designed to help users filtering information,it digs out potential demand,to filter out the items which the users are not interested in,only recommend the items or preference that he or she might like.Collaborative filtering technology is the most used and successful in recommendation system.It usually contacts a user with a group of users who have the same interests and hobbies,then making the specific recommendations.But,in the face of new user cold start and data sparsity,the collaborative filtering recommendation system is still the co-action and facing enormous challenges.To solve these problems,this article puts forward that a combined users attribute characteristic and tree structure of collaborative filtering recommendation algorithm.Joining into user attributes when the similarity calculation,including age,sex,occupation,etc.And then constructs a kind of redundant project subspace to show different interest model,and the model computed based on the similarity of user ratings and user attributes similarity combined according to certain weight,find the target users of neighbors.At last,an improved method named WABR to predict the score of the project which does not have the score according to neighbor users.Then the projects which in the first N are recommended.Through contrast experiment on the Movielens data sets,confirmed that the proposed algorithm can ease the data sparseness problem due to lack of data and new user rating information of cold start problems at a certain extent,eventually making recommendations to a certain extent to improve precision.