Mining Individual Behavior Pattern Based on Semantic Knowledge Discovery of Trajectory
Mining Individual Behavior Pattern Based on Semantic Knowledge Discovery of Trajectory
复制标题
DOI：
10.2498/cit.1002578
复制
发表时间：
2015-08
期刊：
影响因子：
--
通讯作者：
Min Ren;Feng Yang;Guangchun Zhou;Haiping Wang
中科院分区：
文献类型：
--
作者：
Min Ren;Feng Yang;Guangchun Zhou;Haiping Wang
本文试图从原始用户的行为轨迹数据中挖掘出隐藏的个体行为模式。在DBSCAN的基础上，提出了一种新的时空数据聚类算法--基于速度的聚类算法，用于发现用户停止较长时间的单一轨迹的慢速子轨迹(即停止)。该算法使用最大速度和最小停车时间来计算停靠点，并引入分位数函数来估计参数的值，实验结果表明该算法比DBSCAN算法和某些改进的DBSCAN算法更有效、更准确。此外，将停靠点与具有信息呈现特征的POI关联后，设计了POI-行为映射表，根据停靠时间和访问频率分析用户的活动，并在此基础上从历史轨迹中挖掘用户的日常规律行为模式。最终，LBS运营商能够根据个人行为的特点提供智能化和个性化的服务，从而实现精准营销。
This paper attempts to mine the hidden individual behavior pattern from the raw users’ trajectory data. Based on DBSCAN, a novel spatio-temporal data clustering algorithm named Speed-based Clustering Algorithm was put forward to find slow-speed subtrajectories (i.e., stops) of the single trajectory that the user stopped for a longer time. The algorithm used maximal speed and minimal stopping time to compute the stops and introduced the quantile function to estimate the value of the parameter, which showed more effectively and accurately than DBSCAN and certain improved DBSCAN algorithms in the experimental results. In addition, after the stops are connected with POIs that have the characteristic of an information presentation, the paper designed a POI-Behavior Mapping Table to analyze the user’s activities according to the stopping time and visiting frequency, on the basis of which the user’s daily regular behavior pattern can be mined from the history trajectories. In the end, LBS operators are able to provide intelligent and personalized services so as to achieve precise marketing in terms of the characteristics of the individual behavior.