Exit Traffic Flow Prediction Of Toll Station Considering The Spatio-temporal Characteristics Of Road Network

As a key technology of intelligent transportation,traffic flow prediction plays an important role in the field of intelligent transportation system.Accurate and timely traffic flow prediction results can not only allow travelers to have a good understanding of the traffic situation,but also provide a basis for traffic management.The existing research on toll stations mainly uses regression method to analyze the trend of their own traffic changes,and then extrapolates the trend to complete the prediction.It ignores the spatial and temporal connection between toll stations and had a poor performance.Therefore,it is of great significance to predict the traffic flow of toll stations accurately and timely.In this paper,by analyzing the time correlation of charging data,a gated recurrent unit(GRU)flow prediction method combined with time series decomposition is proposed.Based on the analysis of the interaction between toll stations,a flow prediction method considering the time-varying flow contribution between OD was established.Finally,considering the temporal correlation and spatial correlation,a combined prediction method was established to complete the combined prediction of the toll stations flow.The main contents include:GRU traffic flow prediction method combined with time series decomposition.Most of the existing traffic prediction methods for toll stations adopt the method of time series regression and ignore the problem of time correlation between traffic time series.In this paper,a GRU prediction model combined with time series decomposition is proposed to mine the internal correlation of data and complete the traffic flow prediction of toll stations.The experimental analysis of toll data of chongqing expressway is carried out and compared with the traditional differential comprehensive moving autoregression(ARIMA)prediction method and support vector regression(SVR)prediction method.The results show that the proposed GRU prediction method combined with sequence decomposition has higher prediction accuracy.A flow prediction method considering the variability of OD flow contribution.Most of the existing researches ignore the input scale and computability of the models.In this paper,by analyzing the flow distribution characteristics in the road network,determining the optimal dimension of the OD matrix prediction model,reducing the input scale of the model,and then considering the change of the flow contribution between the toll stations,and screening the relevant toll stations reasonable,and the flow transfer weight of each entrance toll station to the target toll station is determined,and then a flow prediction method considering the variability of OD flow contribution is established.The experimental analysis was carried out with the expressway toll data and the real-time data of chongqing expressway.Experimental results demonstrate the validity of the proposed method.A combined prediction method considering spatio-temporal correlation.Most of the existing toll station traffic prediction models are single prediction models,which can only reflect part of the information of the sequence,so they have obvious limitations.In this paper,the dynamic weighting method is used to dynamically determine the weight of each single prediction model according to the prediction performance of the single model in different periods,and the variable weight combination prediction model is established to achieve the effective prediction of the toll station traffic.Using the expressway data of chongqing to carry out the experimental verification,the experimental results verify the validity of the combined forecasting method of toll station flow considering the spatiotemporal characteristics.