Research On Short-term Traffic Flow Prediction Of Road Network Based On Combined Model

The level of urbanization in today's society is constantly improving,and the number of vehicles is also growing rapidly.A large number of cars enter people's daily lives,and they also bring problems such as traffic congestion and pollution.In the face of the increasing traffic demand and the current state of road conditions,the concept of intelligent transportation system was proposed in the early 20 th century,which is a method to effectively solve related problems in the transportation field.In the intelligent traffic management system,the traffic flow prediction algorithm is the key to realize the intelligent traffic management system,it not only affects the control and induction of traffic flow,but also the key to the system from passive response to active control.After analyzing and studying existing theoretical methods,this thesis uses the spatio-temporal,nonlinearity in traffic flow data,and network characteristics of the road structure,relevant research work on single road and road network short-term traffic flow prediction.The main results of the thesis are as follows:(1)Short-term traffic flow prediction research: It is difficult to comprehensively analyze complex characteristics in traffic flow data for a single model,this thesis proposes a combined prediction model-SARIMA-RF model.From the perspective of time-space correlation of traffic flow,the flow data is divided into periodic parts with obvious trends and stochastic fluctuations.Finally,the advanced and effective of the model is shown by comparative experiments.(2)Road network data compression method research: Because the traffic data in the road network is huge,the traffic prediction of all the detection points will lead to the waste of computing resources,thereby affecting the real-time performance of the prediction system.In response to this problem,this thesis proposes a road network data compression method based on MI and CX decomposition.A representative detection point is selected by using the standardized MI value and the grouping threshold ?,and finally constructed as a compression matrix C with a small number of columns.(3)Short-term traffic flow prediction research on road network: The network structure formed by multiple road sections in the road makes the traffic flow data show complex road network spatiotemporal correlation and nonlinearity.Therefore,this thesis proposes DCGRURF short-term traffic flow prediction model.Modeling the traffic flow into a diffusion process on the directed graph,using the DCGRU network captures the spatio-temporal correlation and learns the potential features in the traffic flow data.Finally,RF model derives traffic predictions.(4)According to the traffic forecasting demand,the overall architecture,functional structure and data table of the road network short-term traffic flow prediction system are designed.The above model method is used to obtain the prediction result and visualize the display.The experimental results show that the proposed prediction model is superior to other models in prediction accuracy.The road network data compression algorithm proposed in this paper can effectively simplify road network structureand reduce the calculation difficulty on the basis of ensuring the prediction accuracy.