基于GPS轨迹的用户移动行为挖掘算法.pdfVIP

第 32 卷第 1 1 期 计算机应用与软件 Vo l. 32 No . 11 201 5 年 II 月 Computer Ap p lications an d Software Nov. 2 01 5 基于 GPS 轨迹的用户移动行为挖掘算法 肖艳丽张振字杨文忠 (新疆大学信息科学与工程学院新疆乌鲁木齐 83刷刷 摘 要 挖掘用户的移动行为，可以通过对交通出行方式进行识别来实现。 传统的交通方式识别方法在交通堵塞或多种交通方 式结合的情况下，识别效果并不理想。 针对这种情况，提出基于轨迹分段和监督式学习相结合的识别方法，首先利用速度小于某一 阂值的数据点将原始GPS 轨迹划分为交通方式单一的子轨迹段，然后对子轨迹段分别抽取特征，采用监督式学习方法建立推断模 型对不同子轨迹的交通方式进行识别。实验结果表明，提出的算法能够有效地识别不同交通方式，达到较为理想的效果。 同时在交 通堵塞的情况下也能够很好地识别。 关键词 数据挖掘 GPS 轨迹用户移动行为 交通方式识别特征抽取监督式学习 中图分类号 TP39 1 文献标i只码 A DO I:IO. 3969/j . issn. l000-386x. 2015. 11. 020 USERS MOBILITY BEHAVIOURS 岛mm喝G ALGORITHM BASED ON GPS TRAJECTORY Xiao Yanli Z h ang Z h enyu Yang Wenzhong (ScJwol (扩Informalwn Scie.时e and E咿耐ring， X叼阳Jg Uni.versÎly , Urumqi 83l妙46.Xi.啊ng ,China) Abstract Mining users mobility behaviours can be achieved through the identification of transportation modes. Traditional transportation identification methods do not have ideal effect in traffic jam or the circumstances of various trans阴阳tion modes combination. In view of this situation , we proposed the identification method which is bωed on the combination of trajectory segmentation and supervised learning. First , it divides the original GPS traject町into sub咽jectory sections in single transportation mode using the data points with the speed less than a certain 由reshold.ηlen it extracts 由e features of sub-trajectories 始ctions se阳rately. By building an inference model using the supervised learning me曲od it ident凶es the transportation modes in different s由-trajectories. Experimental 陀sult shows that the proposed algorithm can identify different 田m阴阳tion modes effectively , and achieves rather ideal effect. At 由e same time , this alg