关键词: 用户相似性, 轨迹相似性, 最长公共子序列, 时空挖掘

Abstract: Trajectories do not only record users’ location histories in the physical world but also imply their personal interests and behavior patterns. This paper mainly focuses on the stay region in the trajectory, to find user similarity from the spatio-temporal overlaps of the stay region and propose a method to calculate the user similarity. It detects the stay region, then represents the stay region of the trajectory by using minimum bounding boxes, a similarity measure formula is proposed based on the Longest Common Subsequence algorithm（LCS）. An experiment is performed to evaluate the performance of the new similarity measure by using the trajectories of 60 users in a period of 6 weeks, the results show that the method has high accuracy.

Key words: user similarity, trajectory similarity, longest common subsequence, spatio-temporal data mining