Research On Short-term Traffic Flow Forecasting Method Based On Spatio-temporal Correlation Of Road Network

Traffic signal control and traffic flow guidance technology are the key technologies of intelligent transportation system(ITS).Short-term traffic flow prediction information can provide decision support for advanced traffic information system(ATIS)and traffic management and control system(ATMS)in order to provide travelers Provide effective travel information to improve urban traffic management and control.Therefore,obtaining real-time and accurate traffic flow prediction information is the most critical link and foundation.Many scholars have done a lot of research on the prediction of short-term traffic flow.Among them,through the combination of nonlinear theoretical models,mathematical statistical models,intelligent prediction models and different models,based on the historical data of a certain section of traffic flow to change Most of the prediction analysis is performed,and the spatial and temporal information of the data is not fully utilized.This paper presents a prediction method of RNC-LSTM traffic flow data based on multi-agent graph theory,which combines the spatiotemporal information of traffic flow to improve the prediction accuracy of short-term traffic flow.First,from the time domain perspective,three models of time series autoregression,BP neural network and random forest were used to predict and analyze the short-term traffic flow information of single-segment and multi-segment;Then,from the perspective of space,fully consider the spatial characteristics of urban road traffic flow,use graph theory and multi-agent system theory to build a complex urban road network topology map,combined with iterative learning algorithm,respectively single-segment and multisegment short-term traffic flow The information was predicted and analyzed;Finally,based on the above research,a long-term and short-term memory network(RNCLSTM)traffic flow data prediction method based on road network correlation is proposed.From the perspective of abstract road network correlation,the correlation coefficient matrix is calculated to obtain the model time and space.Information is input and fused to form the input of the prediction system,and short-term prediction of traffic flow is made through iterative learning of multiple agents.Through the analysis of simulation experiment results,the RNC-LSTM model has good performance in short-term traffic flow prediction and analysis.