社交网站中潜在好友推荐模型研究

Latent Friend Recommendation in Social Network Services

社交网站的快速发展深刻地影响了人们的信息共享与交流方式。作为开放的用户交互平台，社交网站的成功很大程度上取决于用户的交互程度和用户黏性。然而，随着社交网站用户规模的爆炸性增长，准确定位兴趣相近的潜在好友对普通用户来说变得越来越困难。本文研究基于用户交互网络的好友推荐方法，分别提出两阶段推荐模型和基于信任传播的推荐模型，通过向用户推荐其可能感兴趣的潜在好友，帮助用户扩大其朋友圈子，进而提高用户黏性。最后，通过对来自Yahoo! Answers和Metafilter两种不同社交网站的用户网络的实验分析，验证了文中推荐方法的有效性。

The rapid adoption of social network services （SNS） has deeply influenced information sharing and interaction process among users. As an open platform for constructing and maintaining social relations among users, the success of SNS websites greatly depends on the quality of user interactions and user viscous. However, as members growing explosively in SNS, it is becoming increasingly difficult for general users to find potential friends of similar interests. In this paper, we focus on recommending latent friends for users based on their interactions. We present a two-stage recommendation model and a trust propagation based recommendation model, respectively. By recommending to a user the latent friends she is interested in, it may help widen her circle of friends, and enhance user viscous accordingly. With experimental analysis on real-world user friendship networks from Yahoo! Answers and MetaFilter, respectively, we empirically evaluated the efficacy of the proposed methods.