Research On Traffic Flow Prediction Method Based On Spatio-temporal Correlation Characteristics

As an important way to improve road safety and alleviate urban traffic pressure,intelligent transportation system has played an increasingly important role.Traffic flow prediction is the basis of intelligent transportation system.Accurate traffic flow prediction is very important for the development of intelligent transportation system.The randomness,nonlinearity and spatio-temporal correlation of traffic flow make accurate prediction of traffic flow still a challenging task.In the face of the complex spatio-temporal correlation and dynamics of traffic data,how to accurately and quickly predict traffic flow so that users can better design their own travel plans has become one of the hot issues in the field of intelligent transportation.In view of the above problems,this paper mainly studies from the following two aspects :(1)Aiming at the problems of local information being ignored in traditional traffic flow prediction and incomplete spatial information capture caused by only relying on the given topological map to capture spatial information,a traffic flow prediction model based on local information enhancement,Spatio Temporal-Convolution Transformer(ST-CT),is proposed.The model can capture both global and local hidden information in the process of traffic flow prediction.At the same time,it does not rely on the given road topology map but relies on the relative position of the road to capture the hidden spatial information and better capture the spatial correlation of traffic data.The experimental results show that the prediction accuracy of the proposed method is improved compared with baseline methods,which proves the effectiveness of the proposed method.(2)Aiming at the problems of road specificity being ignored in traffic flow prediction and the inability to capture other information while capturing time information,a traffic flow prediction model based on data information enhancement-Enhancement Information Graph Recursive Network(EIGRN)is proposed.This method can not only capture the global and local hidden information in the process of traffic flow prediction at the same time.In the capture of spatial information,the topological graph information can be learned through graph embedding and the hidden relationship of each node can be learned by using the adaptive matrix generated by graph embedding,which makes up for the shortcomings of traditional methods in capturing spatial information.At the same time,the personalized expression of traffic topological graph nodes can be enhanced by determining the parameters of each node,so as to better capture the spatial correlation of traffic data,and learn the spatial correlation of data while learning the temporal correlation.The experimental results show that compared with the baseline method,the prediction error of the model is significantly reduced,which proves that it has better effect in traffic flow prediction.