Study Of Short-term Traffic Flow Prediction Based On Deep Learning

With the continuous advancement and acceleration of China's modernization,the research on the prediction of road traffic in expressways has also gradually progressed.The problems and hazards caused by traffic congestion not only affect the pedestrian time,but also increase the environment due to the increase in travel time.This may cause great damage to the economy.Therefore,the prediction of traffic flow is a very important research in urban traffic research.The main contents of this article are as follows:In recent years,researches on short-term traffic flow prediction have mostly adopted LSTM networks based on time-series features.To this research,it is the main disadvantage of this method that a single feature is proposed.The characteristics of the traffic flow are not single features.The spatial characteristics of the road network and the characteristics of the traffic data itself are also particularly important in traffic flow prediction.This paper proposes a short-term traffic flow prediction based on deep learning.First,three types of deep learning neural networks are used to extract multiple features of the traffic flow,including Long Short-Term Memory networks to extract time features,and convolutional neural networks to extract spatial features,and SEA network to extract digital features of traffic flow data.Then fuse the features and predict traffic flow.In the experiments,real-world British highway traffic data was collected,and the data set was pre-processed with missing value filling and normalization to improve the accuracy of the prediction.The validity of the proposed model is verified by comparing the mean absolute error(MAE)and mean square error(MSE)of the proposed model with other models.The experimental results show that the traffic flow prediction method proposed in this paper has greatly improved the accuracy of traffic flow compared with the use of single features,which provides a good support for alleviating traffic congestion.