基于非参数方法的交通流变点问题分析  

Traffic Flow Change Point Analysis Based on Nonparametric Methods

作　　者：王旭琴[1] 胡尧[1] 谌业文 

机构地区：[1]贵州大学理学院,贵州贵阳550025

出　　处：《数学的实践与认识》2016年第12期209-219,共11页Mathematics in Practice and Theory

基　　金：全国统计科学研究项目(2014LZ46);贵州省自然科学基金项目(黔科合J字[2014]2058号);贵州大学研究生创新基金项目(研理工2015056)

摘　　要：考虑到交通流数据的分布形式不确定及其变点数目未知的实际情形,以基于非参数方法的交通流变点问题为研究对象,拟通过基于Kolmogorov-Smirnov(KS)检验和Mann-Whitney U检验的滑动窗口法实现变点存在与否的检验,进一步结合二分法对交通流数据变点数目及其位置进行估计.正态分布模拟仿真显示,两种方法对于均值变点检验和估计效果较好,而对于方差变点检验和估计,Mann-Whitney U方法不及K-S方法.最后,贵阳市中心道路车流量数据实例分析,表明方法对于交通流突变分析效果较好,可为相关部门提供可靠的决策依据.Considered the actuality of traffic flow data with uncertain distribution type and unknown change point number, our research objective is traffic flow change point problem based on nonpaxametric method. We will test the exist of change point by sliding window method based on Kolmogorov-Smirnov（K-S） and Mann-Whitney U tests, and further the location and number of change point will be estimated combing the method of bisection. Normal distribution simulation shows that the two methods are good for the mean change point test and estimation, while for the variance change point, the Mann-Whitney U is worse than K-S method.Lastly, the instance of traffic volume data of the central roads of Guiyang City shows that the effectiveness to traffic flow change point analysis, and could provide reliable basis for related departments.

关 键 词：车流量变点 多变点 K-S检验 Mann-Whitney U检验 滑动窗口法 

分 类 号：O212.1[理学—概率论与数理统计]