交通流信息挖掘的非参数概率变点模型研究  

Study on Nonparametric Probability Change-point Model for Traffic Flow Exploitation

作　　者：王晓原[1] 张敬磊[1] 

机构地区：[1]山东理工大学交通与车辆工程学院,淄博255049

出　　处：《武汉理工大学学报（交通科学与工程版）》2010年第4期801-804,共4页Journal of Wuhan University of Technology（Transportation Science & Engineering）

基　　金：山东省自然科学基金研究项目(批准号:Y2006G32);山东省社会科学规划研究项目(批准号:04CMZ08);山东理工大学科研基金重点项目(批准号:2004KJZ02)资助

摘　　要：交通流状态辨识是智能运输系统(ITS),尤其是其先进的交通管理系统(ATMS)和先进的出行者信息系统(ATIS)研究的一个重要内容.以往算法主要集中在交通流预辨识(即交通流预测)和交通流实时辨识(即事件检测或交通流突变检测)的研究上,而忽略对交通流关键参数量变模式的辨认.依据交通流理论,结合非参数变点统计方法,以在淄博市人民路上采获的拥挤车流数据的研究为例,建立对交通模式量变识别的非线性概率变点模型,对变点存在与否进行假设检验,并对变点(时间)位置的搜索算法作了研究.结合实地数据对本方法进行了标定和检验,验证了本方法的有效性和可行性.Traffic flow condition recognition is one of the important issues for Intelligent Transporta-in advance（namely traffic flow forecasting）and real-time recognition（namely incident detection ortion Systems research,especially for Advanced Traffic Management Systems and Advanced Traveler Information Systems research.The former algorithms mainly concentrated on traffic flow recognition point,a nonlinear probability change-points model to identify quantitative change is firstly establishedwith congestion flow of Renmin Road of Zibo city as an example,the hypothesis-testing problem oftraffic flow breakdown detection）,but recognition of quantitative change of key parameters of traffic flow was neglected.Based on traffic flow theory and nonparametric statistical method of change-change-points and the algorithm to search the change-points are discussed.Finally,the method is calibrated and tested with the field data to verify the validity and the feasibility of the theory.

关 键 词：交通流 量变 变点统计理论 模式识别 信息挖掘 

分 类 号：U491[交通运输工程—交通运输规划与管理]