Research On Road Network Traffic State Recognition Based On Improved K-means Algorithm

With the continuous development of China’s social economy and the increasing number of motor vehicles,the traffic congestion problem in the urban road network is becoming more and more prominent.How to identify the traffic status of road network and reasonably carry out urban traffic control is one of the hot spots of current research in the field of transportation.However,the effect of road network traffic status identification is limited by the quality of traffic data and the accuracy of identification methods.Therefore,it is important to study the methods to improve the quality of traffic data and the efficiency of traffic status recognition.The main work of this paper is as follows:Firstly,on the basis of analyzing the characteristics of traffic data acquisition method,microwave detector data and Internet platform data are used as the sources of traffic data.Based on the deep excavation of the inherent law of traffic data,a traffic anomaly data recognition method is proposed.According to different types and quantities of traffic anomaly data,the corresponding repair method of traffic missing data is established.Secondly,a variety of traffic data fusion methods are compared and analyzed from the basic characteristics,advantages and disadvantages,application level,scope of application,etc.It is concluded that BP neural network method is better.Aiming at the shortcomings of slow convergence speed and easy to fall into local minimum in the learning process of BP neural network method,a traffic data fusion method based on improved BP neural network method is proposed.The BP neural network method is optimized by adding momentum term method and adaptive learning speed method.The minimum error square sum(LSE)is used as the evaluation index to evaluate the fusion effect.Then,the percentile method is used to calculate the free velocity,and the ratio of average travel velocity to free velocity is used as the liquidity index.According to the characteristics of traffic parameters and the use ratio,the liquidity index is selected as the index to divide and identify the traffic state.The traffic state matrix of the road network is constructed based on the one-day mobility index of the road network.According to the NMF algorithm(non-negative matrix algorithm),the traffic state matrix of the road network is decomposed,and the time information and spatial information of the traffic state of the road network are separated.The K-means algorithm is improved by using the particle swarm optimization algorithm,and an improved K-means algorithm is proposed to cluster the decomposed sub-matrix,identify the traffic state of the road network,and analyze its temporal and spatial patterns.Finally,taking part of the road network in the main urban area of Chongqing as the research object,the traffic state of the road network is identified,and its time mode and spatial mode are analyzed.The clustering results of different algorithms are evaluated by running time,tightness and contour coefficient.The results show that the improved Kmeans algorithm proposed in this paper can achieve better clustering results.