基于LDA的用户轨迹分析

LDA-BASED SEMANTIC INTERPRETATION OF USERS TRAJECTORY

随着以GPS导航仪和智能手机为代表的智能终端的普及应用,大量用户轨迹数据得以收集。这些轨迹数据背后隐含了丰富的空间结构信息和用户行为规律信息。对其进行深入挖掘有可能发现用户日常的行为规律,这对城市规划、交通管制等应用都具有非常重要的意义。然而从大量轨迹数据中理解用户行为是一件艰难的工作,提出使用狄利克雷指派LDA(Latent Dirichlet Allocation)模型来对用户轨迹进行语义解释。通过LDA模型可以发现轨迹集中的主题区域以及热门路径,从而可以帮助理解用户的出行意图。实验结果表明LDA能有效地解释用户轨迹。

With the popularity of smart terminals such as GPS navigation devices and smart phones, a large number of users＇ trajectories can be collected. There are rich spatial structure information and user behaviour rules hidden behind these trajectory data. Deeply mining them may find daily behaviours patterns of users, which are very important for urban planning, traffic control and other applications. Howev- er, to understand the behaviour of users from large number of trajectories is a difficult task, here we propose to use latent Dirichlet allocation （LDA） model for semantic interpretation Of user trajectory. Through LDA model we can find topic regions and hot routes where the trajecto- ries converged, which can help to understand the travel intentions of users. Experimental result shows that the LDA can effectively explain user trajectories.