Prediction Of Urban Road Congestion Based On Traffic Situation Data

As an indispensable part of urban transportation,urban road network is an important guarantee for the daily operation of the city.With the advancement of new urbanization in China,urban population and vehicle ownership are increasing,and the increase of road network mileage is gradually becoming stretched.Frequent traffic congestion also reflects the imbalance between network capacity and travel demand.Compared with the traditional traffic data,the status data has the advantages of large volume,wide range and easy access.Therefore,for the management of urban road network,it has great significance to analyze the average speed data of roads in status data and further predict road congestion.In the data preparation stage,the crawler was used to collect three-week traffic status data on the main roads within the second ring of Xi ’ an,which was used as the data basis for the entire article.Then,the wavelet basis function with the best denoising effect was selected by comparison,and the data were denoised by wavelet threshold denoising.Finally,the data was filled in two steps to complete the preliminary data processing.Based on the preliminary data processing,the time and space characteristics of the data were described respectively,aiming to provide a basis for the feature selection of the following model.In the time characteristics,the fast Fourier transform(FFT)was used to determine the main period of the data,the correlation index was used to analyze the correlation and difference between each day of a week,and the peak period of the data was identified.In the space characteristics,the topological relation matrix of roads was established by analyzing the connection between roads,and the correlation between the connected roads was calculated and discussed.Finally,a combination forecasting model was established.Based on the distribution characteristics of the data,the data was standardized.From the perspective of the average road speed,a set of congestion evaluation standards was established which is applicable to both high and low grade roads.Taking into account the chaotic characteristics of the transportation system,the phase space reconstruction of the data was completed,and the reconstructed data was input into the BP neural network to construct a combined model.By using the collected status data to construct a training set and a validation set,the parameters of the model were adjusted and trained.The trained model was used to predict the average road speed,and the predicted value was compared with the congestion evaluation standard to obtain the predicted congestion status of the road.The Lyapunov exponent of the reconstructed data is greater than 0,indicating that the data has chaotic characteristics and the reconstruction is effective.The prediction results show that the combined model has achieved good prediction results.The comparison between different models also prove that wavelet denoising and phase space reconstruction can improve model prediction accuracy.Finally,through the comparison of model accuracy under different delayed prediction times,the applicable prediction time range of the model is obtained.