节点文献
基于位置的社交网络好友推荐算法研究
Research on The Algorithm for Friend Recommendation in Location-based Social Networks
【作者】 王晨；
【导师】 赵守香；
【作者基本信息】 北京工商大学 ， 管理科学与工程， 2015， 硕士
【摘要】 社交网络中的好友推荐算法,通过分析用户社交行为为用户推荐潜在好友,其不仅可以提高好友推荐精度,帮助用户高效的发现潜在的好友,展开社交行为,而且可以保证用户在社交网络中的活跃性,提高用户粘性。近年来,社交网络与逐渐普及的基于位置的服务逐渐融合,形成了基于位置的社交网络。该类网络中用户与好友的关系建立和社交行为开展与地理位置有很强的相关性。但传统的好友推荐算法在产生推荐结果的过程中,往往忽略了位置信息。本文首先结合基于位置的社交网络Brightkite的数据进行挖掘,分析用户使用社交网络的签到功能的行为习惯,并发现用户位置信息对其好友拓扑结构的影响。然后,根据该影响对经典的传统推荐算法进行改进,通过实验进行好友推荐,从而论证改进后的算法,在不大幅提高算法时间复杂度的前提下,可以利用位置信息提高好友推荐结果的精确度。最后,进一步挖掘用户签到的时间信息对好友关系的深一层影响,并尝试将时间信息引入到利用位置信息的算法改进中,进一步提高好友推荐算法的推荐精度。
【Abstract】 Friend recommendation algorithm, as one of the main means to recommend potential friends for user in social networks, which can improvement the precision of friend recommendation, help users to develop their social behavior with potential friends, to ensure the user’s activity in the social network and improve the user viscosity. In recent years, the social network service has integration with Location Based Service which growing popularity, and formed the Location Based Social Network. The user’s relationship with friends and social behavior has a strong correlation between geographical positions. But the traditional friend recommendation algorithms ignore the location information in the process of produce recommendation results.This paper mining the data of Brightkite, one of the Location Based Social Network,analyze the user behaviors of check-in and found the user location information’s impact on friend network topology. Then, according to the influence, this paper gives an improvement to the traditional classic recommendation algorithms. And it demonstrate the improved algorithms can improve the accuracy of friend recommendation results using location information by friends recommend experiments, without significantly increase the algorithms’ time complexity at the same time. Finally, it further mining the check-in time information’s deep influence on the relationship between friends. And it tries to introduce time information into the algorithms’ improvement, to exaltation the accuracy of friends recommendation algorithms.