Urban Regional Traffic Flow Forecast Based On CNN And LSTM

In recent years,a series of traffic problems have emerged as traffic demand increases,especially traffic congestion.In order to effectively alleviate this problem,Intelligent Transportation System(ITS)is widely used in dynamic traffic management.The problem of predicting traffic flow has always been the major project of Intelligent Transportation System.If it is possible to predict the movement of vehicles in an area,then we can prevent and manage it by taking emergency measures.At the same time,the problem of traffic flow prediction is with great challenge,because it is affected by many complex factors,such as inter-regional traffic,events and weather.Therefore,integrating data resources in various fields,improving the allocation of transportation resources and road traffic efficiency,ensuring traffic quality,and the safety of transportation systems have become urgent problems to solve.Under such background,this paper proposes a method that is based on deep learning for predicting traffic flow according to the characteristics of traffic data.First,comprehensively summarize and compare the existing traffic flow forecasting methods.Secondly,elaborate the basic definitions and parameters that are involved in traffic flow prediction.Thirdly,introduce the machine learning and deep learning theory in detail,which establish a theoretical foundation for the establishment of the model.Then,use CNN and LSTM in deep learning as the basic network structure,and combine them to get a new model to predict traffic flow.Finally,use the 2015 taxi trajectory data from Beijing to train and test the model.By comparing with other methods,it is said that the root mean square error(RMSE)of the traffic flow prediction model based on the combination of CNN and LSTM is lower.The root mean square error is 20.03.It shows that the combination of CNN and LSTM can effectively capture spatial features and temporal features in traffic data and accurately predict them.