摘要
随着获取位置信息越来越方便,实时记录用户的移动轨迹成为可能,但是对用户轨迹数据的分析一直停留在轨迹聚类上,而对通过用户的位置轨迹信息分析用户的相似性的研究则较少。为此,提出了分层级多粒度,在不同的邻域半径下密度聚类的方法;改进了传统的聚类算法,探索根据用户的移动轨迹分析用户之间的相似性度量方法。该方法在不同粒度下观察用户访问各个兴趣区域的时长,进而利用向量空间模型(VSM)计算用户在各个粒度下的相似性,最终以不同权重叠加各粒度下的用户相似性值,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。
As location information becomes more convenient, real-time record the user’s movement becomes possible, but analysis of the data on the user trajectory stay on trajectory clusters？ and analyzing information of user similarity by the user’s trajecctory research is less. This paper presents a hierarchical multi-granularity： under different neighborhood radius density clustering algorithm. The raditional clustering algorithm is improved by explore according to user’s mobile trajectory analysis of the similarity between users. The method under different granularity computes user’s stay time of each interested area, and then uses the vector space model （VSM） computing users at each particle size, Eventually stacking weights in different granularity of user^ similarity, the similarity is gotten between the user behavior in geographical space. Experimental results based on real user data show that this method can effectively identify similar user access user’s trajectory.
出处
《计算机与数字工程》
2016年第8期1523-1527,共5页
Computer & Digital Engineering
关键词
聚类
轨迹
数据挖掘
用户相似性
clustering, trajectory, data mining, user similarity
