GMFS:一种新的基于社会网和地理信息的兴趣点推荐方法

GMFS: New point-of-interest recommendation method based on social network and geographic information

基于地理位置的社会网(LBSNs)吸引了大量用户通过签到来分享他们的社会关系和地理信息。通过签到信息可了解用户对地点的偏好,从而给用户提供更好的推荐,因此在基于地理位置的社会网上进行兴趣点(POI,Point-of-Interest)推荐逐渐成为了热点研究问题。以往的研究没有将社会关系和地理信息联合融入到基于矩阵分解的POI推荐方法中。基于社会网和用户签到活动日志,提出了用加权的方法计算用户之间的相似性,在此基础上提出了一个联合社会网和地理信息的加权矩阵分解模型GMFS,并给出了高效的求解方法。多个真实数据集上的实验结果表明:GMFS方法能有效地进行POI推荐。

Location-based social networks(LBSNs) have attracted millions of users to share their social relation ship and their locations via check-ins. The available check-in information makes it possible to mine user’s preferences on locations and provide better recommendations for users, and thus POI(Point-of-Interest) recommend on LBSNs has became a hot in research work. However, previous works do not investigate the POI recommendation method that fuses social network and geographic information into matrix factorization. In this paper, a weighted method for computing the similarity between users based on social network and user check-in activity log was proposed. Based on the weighted method, a weighted matrix factorization model GMFS that combines social network and geographic information was proposed, and present an efficient method for estimating GMFS parameters. Experimental results on several real datasets show that the GMFS method can effectively recommend POIs for users.