LBSN中考虑用户交友偏好的好友推荐方法研究

基于位置的社交网络(location-based social networks,LBSN)大为流行之余,也带来了信息过载问题.好友推荐是所有社交网络必须面临的问题,为了改进LBSN中好友推荐的效果,构建了考虑用户交友偏好的好友推荐模型(friends recommendation considering users'preference,UPFR).从兴趣相似性、距离和熟识度三个属性刻画LBSN中的用户,兴趣相似性属性基于信息熵理论计算、距离属性通过朴素贝叶斯推导、熟识度属性建立在共同好友的基础上.在对三个属性进行集成时,考虑了用户的交友偏好,通过目标用户的好友列表确定各属性的权重,建立了自适应用户交友偏好的好友推荐算法.通过Foursquare上的数据实验证明该算法能取得较优的综合推荐效果.

A study of friends recommendation algorithm considering users' preference of making friends in the LBSN

With the popularity of location-based social networks (LBSN), information overload appears. Friends recommendation has been the problem related with all kinds of social networks. To improve the friends recommendation performance in the LBSN, a kind of friends recommendation algorithm considering users' preference of making friends is proposed. Users in LBSN were described by similarity-of-interests attribute, distance attribute and familiarity attribute. The similarity-of-interests attribute was computed based on information entropy theory. The distance attribute was computed based on Naive Bayesian. The familiarity attribute was computed based on the quantity of same friends. Users' preference was reflected by weighting coefficient of the three attributes which comes from the friends list of the user. By the test on Foursquare, it shows that the algorithm achieves good performance.