Research On The Evaluation And Prediction Of Traffic Conditions At Road Intersection

With the rapid development of the national economy,the number of motor vehicles and people’s travel demand continue to increase,causing traffic congestion.The problem is also becoming increasingly serious.Therefore,it is necessary to vigorously develop intelligent transportation and constantly improve traffic management strategies.Through comprehensive evaluation and accurate and efficient prediction of traffic status at urban road intersections,it can provide scientific decision-making basis for traffic management departments,as well as easily understandable road congestion information for the public.This paper mainly focuses on the evaluation and prediction of traffic status at urban road intersections.The main tasks are as follows:Firstly,evaluate the traffic status of the intersections.Taking into account various traffic index parameters,a multi indexes evaluation system for traffic status is constructed using the literature review method and frequency statistics method to select three indexes: average vehicle speed,saturation,and average vehicle delay time.According to the classification results of traffic status under a single index,an index membership matrix is constructed,and the comprehensive evaluation results of traffic status at each intersection are obtained by calculating the “membership matrix-weight”.The intersection traffic status are determined using the maximum membership principle.For the calculation of index weights,it is expected to obtain a set of weight values that have the smallest deviation from the evaluation results of different subjective and objective weighting methods.This paper has made improvements to the combined weighting model based on the Least Squares(LS)method.Considering that the relative importance of subjective and objective weights varies for different evaluation indexes,a coefficient of relative importance of subjective and objective weights has been added to the model,which keeps the index importance information reflected in the single weighting result to the maximum extent,and further improves the rationality of the weighting results of the model.Secondly,the intersections traffic indexes data are predicted.Traffic data has obvious periodicity and spatiotemporal correlation.The above characteristics are explained in the paper and verified with Pearson correlation coefficient.This paper proposes to establish a prediction model based on traffic data decomposition and spatio-temporal features extraction(TDD-STFE),which improves the accuracy of traffic data prediction by adding the above features information.The model first converts the linear part of the traffic data into a periodic Fourier series,and then subtracts the linear part from the original traffic data to obtain the nonlinear part,achieving the decomposition of traffic data.Different methods are used to predict two parts of data with different properties.The linear part calculates the periodic value of the future time based on the Fourier series expression,and the nonlinear part extracts spatio-temporal features based on the Convolutional Neural Network-Gated Recurrent Unit(CNN-GRU)model to achieve effective prediction.Then,the two parts of the predicted values are added to obtain the final predicted values.In this paper,three traffic indexes are predicted based on real traffic data sets.Compared with other models,the prediction accuracy of the model proposed in this paper is higher.Finally,the predicted values of each index are substituted into the intersection traffic status evaluation model to achieve traffic status prediction.The validation of actual traffic data shows that the evaluation and prediction methods for traffic status at intersections proposed in this paper are both practical and effective,and have certain positive significance for traffic control and guidance at intersections.