Urban Road Traffic Flow Prediction And Traffic State Recognition

With the growth of social economy,urban traffic management is facing increasing bigger and bigger pressure.Rapid and accurate traffic state prediction and identification technology can provide information support for traffic management,help drivers plan travel routes and reduce traffic congestion.Therefore,in order to alleviate road congestion,correctly predict and identify future traffic flow,this dissertation combines the advantages of different prediction models,improves the background value of grey model and combines with BP neural network to predict future road traffic flow.Owing to the limitations of the traditional clustering algorithm,we introduce maximum density and furthest distance criterion to optimize the initial clustering center,classifies traffic conditions by using traffic volume,speed and occupancy,contrast speed threshold method recognition results,verify the feasibility of the recognition algorithm,according to the predicted traffic volume,speed and occupancy identify future road traffic state.In this dissertation,traffic data acquisition methods are summarized and compared,and the data collection methods are determined.Repair the data according to data preprocessing method,we get the final data sequence.From the perspective of managers and travelers,the traffic states are divided into four categories,analyze the influence degree of each parameter on the traffic status,the traffic volume,speed and occupancy are determined as the identification indicators of the traffic state,and the identification system of the traffic state is constructed.In addition,this dissertation presents a weighted grey neural network prediction method.The biggest impact on the traffic flow at the next moment is often the traffic flow data at the previous moments.Grey model can be modeled by using less data,while updating data constantly,remove old data with little significance,and dynamically reflect the latest change trend of traffic flow.However,when the data volatility is large,the prediction accuracy is low,and the BP neural network model can make up for this shortcoming.Therefore,the grey model and BP neural network are combined to predict the future traffic flow,so that the prediction results have the advantages of different models at the same time,then the final prediction results being corrected with weights.Taking traffic volume prediction as an example,the errors of different prediction models are contrasted with examples to verify the feasibility of the combined model.Finally,according to the prediction model,the speed and occupancy are predicted respectively.In view of the limitations of existing recognition algorithms using single traffic index,this dissertation uses traffic volume,speed and occupancy to identify traffic states.The initial clustering center of the clustering algorithm is improved by the sample density and the distance between the samples,the euclidean distance calculation is improved by combining subjective and objective ideas,establish the improved traffic status recognition model,and the performance of the improved algorithm is verified by an example.The speed threshold method is compared with the proposed method,and the results show that the proposed algorithm has some advantages over the speed threshold method.Finally,according to the actual prediction results,input actual road traffic parameters to identify future road traffic status.