基于谱聚类与RS-KNN的城市快速路交通状态判别  

Traffic State Identification for Urban Expressway Based on Spectral Clustering and RS-KNN

作　　者：商强[1] 林赐云[1,2] 杨兆升[1,2] 邴其春[1,3] 田秀娟[1] 王树兴[4] 

机构地区：[1]吉林大学交通学院,吉林长春130022 [2]吉林大学吉林省道路交通重点实验室,吉林长春130022 [3]青岛理工大学汽车与交通学院,山东青岛266520 [4]山东高速公路股份有限公司,山东济南250014

出　　处：《华南理工大学学报（自然科学版）》2017年第6期52-58,共7页Journal of South China University of Technology(Natural Science Edition)

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

摘　　要：为了提高城市快速路交通状态判别的准确性,构建了一种基于谱聚类与随机子空间集成K最近邻(RS-KNN)的交通状态判别模型.以地点交通参数为基础,根据交通流运行特性并结合中国道路服务水平的4个等级,采用谱聚类算法将交通状态划分为4类;然后使用已分类的交通流数据训练RS-KNN模型.通过上海快速路的实测数据完成模型的实验验证和对比分析.实验结果表明,所提出的模型不仅能够提高交通状态判别的精度,而且具有良好的鲁棒性,其判别率比标准KNN模型、BP神经网络模型和SVM模型分别提高7.3%、4.9%和4.5%.In order to improve the accuracy of traffic state identification for urban expressway, a traffic state identi- fication model based on spectral clustering and RS-KNN （Random Subspace Ensemble K-Nearest Neighbors） is de- veloped. In the investigation, first, on the basis of spot traffic parameters data and according to the operation cha- racteristics of traffic flow, the traffic state is divided into four categories with the consideration of the four levels of service for Chinese roads. Then, the classified traffic flow data are used to train the RS-KNN model. Finally, by using the real data of an expressway in Shanghai, China, an experimental verification and a comparative analysis for the proposed model are carried out. Experimental results demonstrate that the proposed model not only improves the accuracy of traffic state identification but also possesses good robustness ; and that the identification rate of the proposed model is 7.3% , 4.9% and 4. 5% higher than that of the standard KNN model, the BP neural network and the SVM model, respectively.

关 键 词：交通工程 交通状态判别 谱聚类 随机子空间 K最近邻 

分 类 号：U491[交通运输工程—交通运输规划与管理]