Location-based social network allows users to share location information. The complete geographical record about users kept by social network plays as the basis for analyzing the behaviors of the users in geographical track. For Location-Based Service (LBS) platform did not take the users' geographical location on the track into consideration, this paper proposed a new hierarchical density based clustering approach. It determined the similarity among users in different scales by classical Vector Space Model ( VSM), with vectors composed of users' visiting frequencies about different cluster area. Overlapping the different scale user similarity value with different weighted obtained the geospatial similarity of the user behaviors. The experiments based on user data from a large LBS social network site demonstrate that the proposed approach can effectively identify similar users.%基于位置的社交网络(LBSN)能够支持用户分享地理位置信息,网站中保存用户访问真实世界地理位置的记录构成用户的行为轨迹,但LBSN用户相似性的分析并没有从用户的地理位置轨迹上加以考虑.为此,提出基于划分层次,在不同的邻域半径下密度聚类的方法,探索基于位置的服务(LBS)平台上用户地理位置上相似性的度量.该方法在不同空间位置比例尺下观察用户访问各个聚类区域的次数,进而利用向量空间模型(VSM)计算用户在各个层级的相似性,最终以不同权重叠加各层级的用户相似性值,得出用户在地理空间行为上的相似性.基于国内某大型位置社交网站真实用户数据的实验结果表明,该方法能有效识别出访问地理位置相似的用户.
展开▼
