Study On Road Network Traffic State Index Based On Multi-source Data Drive

With the rapid development of city,the traffic flow of road network is becoming more and more complex,which makes the traffic state will not only change with the time,but also be affected by spatial factors.The maturity of intelligent detection technology and the widespread application in the field of transportation,more and more types and quantities of traffic data is provided for transportation systems.And fusion processing of multi-source massive data can provide more comprehensive traffic information for traffic state index analysis and traffic flow fine control.Because of the disadvantages of traditional data processing methods based on prior knowledge and mathematical model,such as complicated parameters and excessive assumptions,data-driven intelligent algorithms are used to fuse multi-source massive data to obtain complete and reliable traffic state indicators data.Meanwhile,considering the dynamic influence of physical characteristics,traffic flow changes and spatio-temporal effect on state weights,using indicators fusion data and data-driven technology to dynamically analyze the traffic state index can provide effective information support for traffic management control and residents’ travel choices.Firstly,based on the demand of the traffic state index for static and dynamic data,the performance characteristics of data collection technology and the basic features of multi-source dynamic data are analyzed.And the quantitative indicators of traffic state index,classic calculation models,and common data-driven methods are introduced.Secondly,according to the abnormal situation of the floating vehicle GPS data,data processing and map matching method based on Arc GIS is proposed.On the basis of dynamic threshold and traffic flow mechanism identification for fixed detection(video,microwave,geomagnetic)data,a fault data repair prediction model based on PSO-SVR is constructed.Considering the limitations of single-source data and the complementarity of multi-source information,on the premise of matching time and space,combining with the advantages of global search of genetic algorithm and adaptive learning of wavelet neural network,a multi-source data fusion model based on GA-WNN and least square method is designed.The results of case study show that:high-quality and high-coverage state indicators data can be obtained at a faster speed based on repair and fusion models.Nextly,considering the differences in the importance and influence of links and roads in the regional road network,based on the static physical attributes and dynamic traffic demand of links,and combined with the spatial-temporal correlation coefficient to quantify the influence degree of the traffic flow between the link and adjacent links,the importance evaluation indicators based on multi-attribute is constructed.a model of the links importance degree calculation based on hierarchical-entropy weight TOPSIS decision-making is established.According to this,the traffic state impact weight is dynamically determined based on link importance index and road vehicle kilometers,which provide a basis for the calculation of roads and network traffic index.Finally,targeting the ambiguity and subjectivity of the traffic state,a method of traffic state division based on SAGA-FCM is proposed by using global search technology of genetic algorithm and local optimization ability of simulated annealing algorithm.And the range of state indicators corresponding to different traffic states are determined.According to this,a traffic state index model is constructed based on the intuitionistic fuzzy entropy theory combined with the indicators classification interval and the state impact weight.The dynamic calculation of the state index and the real-time partition of the corresponding levels are implemented from three levels: links,roads and regional road networks.The result of case study show that: traffic state index model of multi-source data-driven has strong stability and high accuracy.