Research On The Content And Social Filtering Algorithm For Friend Recommendation

With the rapid development of the Internet, the human society has officially enteredthe network interactions era of the information society and the Internet. A variety ofdifferent social websites have emerged,such as Facebook, Twitter, Flickr, Sina micro blog,Renren and so on. The most basic function of social network is to provide other users tosocial network users for their friends. In the face of a huge number of social network users,add what users for their friends, has become a huge problem. The social network of friendsrecommended system is a response to this problem. Most existing friends recommendedalgorithm in social networks only consider the long-term use of the user, the accuracy ofrecommending for new registered user is not good enough. Too much to consider mutualfriends between users or recommending excessive some hot users cannot be able toprovide users with personalized friends.This paper analyzes the current popular recommended algorithm detailedly, friendsrecommended mechanism which has been used in the social networks has been researched.At present, to solve the social network friends recommended problems, an algorithm isproposed based on the content and social filtering, this algorithm can be integrated toconsider the user’s personal information and the existing network of friends. Firstly, thispaper gives the definition of objects in a social network, and analyzes the most commonlyused optimization method of recommended algorithm. Secondly, content-basedrecommended algorithm is proposed to calculate the similarity between users, obtained auser set of a highest similarity with the target user. Then, social filtering algorithm isproposed, calculated common friends of the maximum number of users with highestsimilarity in the target user set with association rules, form a set of users with mostcommon friends. Finally, combined with the user’s personal information and socialrelations of recommended factors, personalized friends recommended algorithm is derived.In the experimental section, recommended system of verification algorithm isdesigned, provide the technical platform support for testing the performance of thealgorithm. In the treatment of experimental data, filter out the users which have lesspersonal information and social relationships, test the performance of the personalizedalgorithm. In order to verify the proposed algorithm can also be effective for users whichhave less personal information and social relations, recommended system add a function ofadding new users, re-validate the algorithm with adding the users which have been filtered out. The evaluation accuracy rate and recall rate of experiment show that this methodeffectively solves the problems of the new registered users and less user information whichcannot be accurately recommended in recommendation of friends, meanwhile, comparewith the traditional algorithm, the performance of the algorithm is obviously improved onthe recommendation of friends.