Merging Into User Interest And Trust Friend Recommendation Algorithm Research

With the rapid development of network technology, we gradually ushered in theSocial Network, some famous social networking site has become an influentialplatform for the information, such as Facebook, MySpace, Flickr and so on. In the eraof social networks, many users transfer offline relationships to online relationship.Therefore, the friend recommended function as a social network is very popular anduseful personalized service, and help the customers to establish a good relationshipwith friends more quickly. But due to the sparse data (Data sparseness) and cold startproblem, friend recommend systems are lack of adequate data information, so that therecommendation result is low accuracy and the degree of personalization. How in thefriend recommended system more reasonable mining data information is the key tosolve the above problems. The main research content of this article is as follows.First, an overview of related knowledge of the friend recommended system isintroduced, including the recommended application, model and characteristics of theuser, the development of history and status quo of the recommendation system.Introduces the several common friend recommendation technologies, as well as thetechnology of each specific application areas and the existing problems.Secondly, according to the traditional recommendation algorithm based on User-based, take advantage of user interest model, calculating the interest degree of users.And then improve calculation method of the user similarity, merge into the user’sinterest the algorithm in some extent. Improve the accuracy and personalization of thefriend recommendation. Considering the actual effect of recommender systems, wefurther describe in detail in the user interest after the improved algorithm.Then, considering the trust network playing an important part in socialnetworking sites, we put forward to the user’s social trust integrating into good friendrecommended approach. Through the selection of the trusted neighbor set, weimprove user ratings. Finally on the basis of collaborative filtering algorithms andsimilarity formula by users into the trust, produce the final recommendations.Finally, the two proposed methods in this paper merging into the user’s interests recommendation algorithm and integrating into the social trust friendrecommendation respectively carry on the contrast experiment, and the forecastingaccuracy and diversity of the algorithm is analyzed.