一种面向获取空间信息的潜在好友推荐算法

发布时间：2018-03-31 12:20

本文选题：LBSN(location-based　切入点：mobile　出处：《软件学报》2017年08期

【摘要】：随着社交网络的不断发展,朋友推荐已成为各大社交网络青睐的对象,在能够帮助用户拓宽社交圈的同时,可以通过新朋友获取大量信息.由此,朋友推荐应该着眼于拓宽社交圈和获取信息.然而,传统的朋友推荐算法几乎没有考虑从获取信息的角度为用户推荐潜在好友,大多是依赖于用户在线的个人资料和共同的物理空间中的签到信息.而由于人们的活动具有空间局部性,被推荐的好友分布在用户了解的地理空间,并不能满足用户通过推荐的朋友获取更多理信息的需求.采用用户在物理世界中的签到行为代替虚拟社交网络中的用户资料,挖掘真实世界中用户之间签到行为的相似性,为用户推荐具有相似的签到行为且地理位置分布更广泛的陌生人,能够增加用户接受被推荐的陌生人成为朋友的可能性,在保证一定的推荐精度的基础上,增加用户的信息获取量.采用核密度估计估算用户签到行为的概率分布,用时间熵度量签到行为在时间上的集中程度,选择可以为用户带来更多新的地理信息的陌生人作为推荐的对象,通过大规模Foursquare的用户签到数据集,验证了该算法能够在精度上保证与目前已有的LBSN上陌生人推荐算法的相似性,在信息扩大程度上高于上述已有算法.
[Abstract]:With the continuous development of social networks, friend recommendation has become a favorite of social networks, which can help users broaden their social circle and get a lot of information through new friends. Friend referrals should be aimed at broadening the social circle and getting information. However, traditional friend recommendation algorithms rarely consider recommending potential friends to users from the point of view of getting information. Most rely on online personal data and check-in information in a common physical space. Because people's activities are spatially localized, recommended friends are distributed in the geographic space the user understands. It can not meet the need for users to obtain more information through recommended friends. The user's check-in behavior in the physical world is used instead of user data in virtual social networks, and the similarity of check-in behavior among users in the real world is mined. Recommendation of strangers with similar check-in behaviors and wider geographical distribution can increase the likelihood that users will accept recommended strangers as friends and ensure a certain recommendation accuracy. The kernel density estimation is used to estimate the probability distribution of the user's check-in behavior, and the time entropy is used to measure the concentration of the check-in behavior in time. Select strangers who can bring users more new geographic information as recommendation objects, and sign in data sets through large scale Foursquare users, It is verified that the proposed algorithm can guarantee the similarity with the existing strangers' recommendation algorithm on LBSN in accuracy, and is higher in the degree of information expansion than that of the existing algorithms.
【作者单位】：哈尔滨工业大学计算机科学与技术学院;哈尔滨理工大学软件学院;
【基金】：国家自然科学基金(61370214,61300210)~~
【分类号】：TP391.3

【相似文献】