Weak Ties Recommendation Methods Research In Social Networks

People recommendation on social networks is a topic with great practical significance. Especially today, a variety of social networking applications emerging, people recommendation effectively helps users to filter overload information, thus improves user stickiness to social network applications. However, previous studies about recommendation have focused on the merchandise, movies, music recommendation, which may not be applicable in people recommendation on social networks. People recommendation has its own unique side, research on various online social networks people recommendations is very valuable.This paper argues that people recommendation can be divided into two categories, one is strong ties recommendation which takes aim at the offline social relations rediscovery, another is weak ties recommendation which takes aim at finding new "friends" those users may be interested in. This article will focus on weak ties recommendation. We will introduce the definition of weak ties, discuss the significance and difference of weak ties recommendation, evaluation existing methods, propose our own ideas and verify them with experiments.Since previous recommender systems don’t distinguish th e strength of relationship between the characters, this paper firstly classifies the existing people recommendation method according to the previous survey, select eight kinds of commonly recommendation strategies, compares them on different datasets by th e performance on weak ties mining, figures out the characters of those good algorithm.On the basis of comparative analysis, this paper proposes a novel weak ties recommend method which combines community division and Meta-Path to solve the collaborator recommendation problem on bibliographic networks. The Algorithm uses communities to find similar users to a given user, thereby expanding user features. Experiments demonstrate the effectiveness of our proposed algorithm.Sociology studies have shown that weak ties are often able to bring more diverse information, this paper finally uses a hybird strategy which combine the results of two recommendation algorithm to increase the diversity performance of people recommendation on Microblog. The strategy combine community partition based Meta-Path algorithm and feature based Maxtrix decomposition algothm in detail. The results verify the effectiveness of weak ties oriented recommendation algorithm in enhancing diversity.