摘要
有效的潜在好友推荐是促进社交网络不断增长的重要途径,基于位置的社交网络(LBSN)支持用户随时随地记录自己所处的位置以及分享身边发生的事,用户通过签到生成的地理位置信息构成了用户的行为轨迹。文章提出了一种利用用户签到信息的潜在好友推荐方法,不同于传统的只考虑签到位置信息的方法,文章中的方法还考虑了签到的时间以及签到的次数。方法首先将签到的时间信息及空间信息压缩,然后进一步通过计算用户之间的相似度来决定向目标用户推荐的潜在好友。相似度主要决定于用户之间的重合点数量、近似点数量以及签到点集大小。文章最后使用大型位置服务社交网络Gowalla的真实用户访问数据,以准确率与召回率作为评价标准,证明了本文使用的方法在效果上优于传统的使用位置信息的潜在好友推荐方法。
Effective friend recommendation effective potential is an important way to promote the growing social network. Location-based social networking (LBSN) Supports user to record their location anywhere and share things happening around. Location information generated by the user constitute the user's behavior trajectory. This paper presents a potential friend recommendation method using the location information of user. Different from conventional method which only consider location information, this paper also consider the influence of time and the number of check-in. Firstly, this method compress time information and cluster location information. Then further by the similarity calculation to determine the user's recommendation to target potential friends. Similarity is mainly determined by the number of common point, the number of similar point and the user check-in point set size. Finally, this paper use a large scale dataset which come from location-based social network Gowalla’s real user access record with accuracy rate and recall rate as an evaluation to validate that this method is better than the conventional method.
出处
《软件》
2015年第1期62-66,共5页
Software
关键词
好友推荐
位置服务
用户相似度
签到数据
friend recommendation
location service
user similarity
check-in data
