基于投影寻踪动态聚类的快速路交通状态判别  

Traffic State Identification for Urban Expressway Based on Projection Pursuit Dynamic Cluster Model

作　　者：邴其春[1] 龚勃文[1,2,3] 杨兆升[1,2,3] 林赐云[1,2,3] 曲鑫[1] 

机构地区：[1]吉林大学交通学院,吉林长春130022 [2]吉林大学汽车仿真与控制国家重点实验室,吉林长春130022 [3]吉林大学吉林省道路交通重点实验室,吉林长春130022

出　　处：《西南交通大学学报》2015年第6期1164-1169,共6页Journal of Southwest Jiaotong University

基　　金：国家科技支撑计划资助项目(2014BAG03B03);国家自然科学基金青年基金资助项目(51308248;51408257);山东省省管企业科技创新项目(20122150251-1)

摘　　要：为了提高快速路交通运行状态的判别精度,利用地点交通参数与交通状态之间的映射关系,提出了基于投影寻踪动态聚类模型的快速路交通状态判别方法.该方法综合投影寻踪技术和动态聚类方法构造投影指标函数,采用混合蛙跳算法优化投影指标函数的投影方向获得最佳投影方向,并利用仿真数据标定了交通状态判别阈值.结合仿真数据和实测数据进行了实验验证和对比分析.实验结果表明,投影寻踪动态聚类模型能够有效提高快速路交通状态判别精度,平均判别率为97.01%,平均误判率为0.86%,平均判别精度分别比BP神经网络模型和模糊C均值聚类模型方法提高了8.9%和4.5%.In order to improve the accuracy of traffic state identification for urban expressway based on the spot traffic parameters,a traffic state identification method based on projection pursuit dynamic cluster model was proposed using the mapping relationship between spot traffic parameters and traffic state. First,the projection index function was constructed by combined use of the projection pursuit technology and the dynamic cluster method,and the shuffled frog leaping algorithm was used to optimize the projection direction. Then,the traffic state identification threshold was determined using simulation data. Finally,validation and comparative analysis were carried out using both the simulated data and measured data. Experimental results indicate that the proposed model can effectively improve the accuracy of traffic state identification. The average identification rate is 97. 01% and the average false identification rate is 0. 86%. The average identification accuracy of proposed method is 8. 9%and 4. 5% higher than the BP neural network model and the fuzzy C-means clustering model,respectively.

关 键 词：交通状态判别 投影寻踪动态聚类 混合蛙跳算法 城市快速路 

分 类 号：U491[交通运输工程—交通运输规划与管理]