Research On Traffic State Identification And Decision-making Model Of Highway By Fuzzy Clustering

At present,a large number of data collection devices such as different scale of TV monitoring system and different specifications of vehical detector have been installed in most of the highways in china.But the massive data that collected by these devices was not fully used resulting in identification of traffic state generally staying manual evaluation stage.For this reason,based on the analysis of the characteristics of massive traffic data,this paper constructs a traffic state division criterion and a real-time decision model for a certain highway which is of great significance to improve the management efficiency of highway.Fist of all,the index of the traffic state and their relationship were analyzed.And it is arccording to the principle of index selection and the advantages and disadvantages of data acquisition technology,the index used in traffic indetification are determined.Then,the fault recognition and corresponding processing method of traffic data are given based on the limitation of the data acquisition devices.At the same time,the charcateristics of traffic data are analyzed from three aspects: time series trend,mutual influence and probability statistics.After that,on the basis of the analysis of the division standard in domestic and foreign,this paper puts forward a traffic state classification method by fuzzy clustering and use fuzzy c-means(FCM)clustering algorithm to analyze historical traffic data.Afterwards,taking the clustering center as the critertion basis and combining with the characteristics of traffic data,a Comprehensive identification method by combination with single-index identification and multi-index identification was proposed.Furthermore,the two-level alarm model of congestion decision and qualitative analysis method of describing the traffic state of sections are put forward.In consideration of the shortcoming of the FCM algorithm that depends on the initialization parameters,parameters including the data set,the weighted index,the chuster number and the initial clustering center are optimized and accordingly the FCM algorithm is improved.Finally,taking the traffic data from kaiyang highway,Guangdong Province,for case study,it is verified that the traffic state identification method and decision-making model are effective and feasible.