Traffic Flow Forecasting Method Based On Temporal And Spatial Association Model

Traffic flow forecasting is a fundamental issue in traffic modeling and traffic management.In recent years,although the method of deep learning has been applied to the field of traffic flow prediction,due to the characteristics of the traffic data itself,such as spatial dependence,time dependence,and external factors,accurate prediction of traffic flow is still faced great challenge.This article mainly studies the prediction of traffic flow.It mainly divides a city into different grids.Each grid represents a region.Then this grid is used to predict the traffic flow of the entire city.Aiming at the temporal-spatial relationship characteristics of traffic flow data,this paper proposes a new hybrid model prediction method based on deep learning,using traffic convolutional neural networks(CNN)and LSTM networks that integrate Attention mechanism to carry out traffic flow.Predict and verify our experimental results with experiments.The main research work of this paper is as follows:(1)By dividing the map of a city into grids according to latitude and longitude,the prediction of urban traffic flow can be successfully translated into problems in the field of image processing,and the deep learning framework can be used to deal with related issues.The extensive application of deep learning in the field of image processing makes it possible to better extract relevant features of traffic data.(2)Aiming at the spatial-temporal characteristics of traffic data,this paper proposes a new framework of deep learning LSTM based on convolutional neural network and recurrent neural network.At the same time,the Attention mechanism and autoregressive model are added to the framework.To more accurately predict traffic flow.CNN is mainly used to obtain the spatial relationship between traffic data.LSTM can obtain long-term dependencies of time series.Attention mechanism can better focus on the local characteristics of time series.Autoregressive model can be used to obtain time series.The linear relationship between.(3)Different experiments were used to verify the role of each module in the overall framework,that is,to verify its effectiveness and necessity.Afterwards,through the corresponding experiments,the parameters of each module used in the framework of this paper were selected.Finally,through comparison experiments,we found that our experimental results can achieve better results than previous methods for predicting traffic flow.