Research On Multi-step Traffic Flow Prediction Based On Deep Learning

In recent years,with the accelerating process of urbanization,traffic congestion has caused economic,social and environmental problems in many cities around the world.As an important part of the new generation of intelligent traffic management system,accurate real-time traffic prediction can not only enable traffic managers to grasp the traffic conditions and traffic conditions of the road in time for a period of time,and plan and guide traffic in advance,but also provide travel suggestions for road travelers to ensure better mobility and less congestion.It is also of great significance to the road planning and traffic management of smart cities in the new era.With the extensive use of traffic detectors and sensors on urban road networks,modern transportation systems have accumulated a large amount of historical data,such as flow,speed and density.As a new method,deep learning can dig out the inherent characteristics and rules of data from massive data by optimizing and training the network,which has strong data learning ability and feature expression ability.Therefore,this paper applies the deep learning method to traffic flow prediction.The main research contents of this paper are as follows:1.Aiming at the traffic flow prediction of traffic section,a multi-step traffic flow prediction method based on Conv1D+LSTM is proposed in this paper.The method combines onedimensional convolution Conv1 D with long short term nemory network,and external factors affecting traffic flow prediction such as weather,holidays and time information were taken into consideration.Conv1 D was used to model the time characteristics,period characteristics and external related characteristics of traffic flow data,and then the extracted features were sent to LSTM for multi-step traffic flow prediction.On the basis of ensuring the accuracy of traffic flow prediction,this method can also help people better understand the influence of external factors on traffic flow prediction.2.Aiming at the traffic flow prediction of traffic network,a multi-step traffic flow prediction method based on AST-GCN-LSTM is proposed in this paper.This model uses attribute enhancement unit to extract external features(weather,POI,time,etc.)that affect traffic flow.Local Spectral Convolution(LSGC)is used to capture spatial correlation features from K-order local neighbors of each road node in the road network.LSTM is used to capture the time dependence of traffic flow sequences.The experimental results show that the proposed model is not only suitable for the traffic flow prediction of the road network,but also suitable for the medium and long term traffic flow prediction,and the prediction accuracy of the ASTGCN-LSTM model is significantly higher than the benchmark model.