Research On Short-term Traffic Flow Forecasting Method Based On Machine Learning

Recently,transportation demand is growing and a series of traffic problems have emerged,which the most obvious is traffic congestion problem.In order to alleviate this problem effectively,Intelligent Transportation System(ITS)is widely used in dynamic traffic management.As the important basis of the traffic control,short-term traffic flow has always been the focus of Intelligent Transportation System(ITS).Accurate short-term traffic flow forecast can not only realize the dynamic control of traffic state and implement traffic guidance,but also provide real-time and concrete road information for travelers and improve road capacity.With the continuous breakthrough of big data and information technology,short-term traffic flow collection methods have been improved and more and more data are produced,so how to use the massive traffic data for more accurate short-term traffic flow forecast has become an urgent solved problem.In the background,this paper proposes a short-term traffic flow forecasting method based on machine learning,which uses the advanced deep learning of the machine learning field to predict the short-term traffic flow.First,it is in view of the shortcomings of the current shortterm traffic flow forecasting model.Secondly,the basic parameters,characteristics and acquisition methods,pretreatment and so on of the short-term traffic flow are analyzed in detail.Thirdly,deep learning theory is discussed,which lays the foundation for the follow-up prediction.Then,the LSTM model is used to forecast the short-term traffic flow,and the network structure,training process,parameter selection and concrete steps of the model are described.Lastly,the short-term traffic flow in Changde City and the short-term taxi volume of Xi'an are used to validate the model.By comparing with other models,the MAE and MAPE of the RNN short-term traffic flow prediction model based on LSTM are lower than others and the prediction accuracy is about 93%,which indicates that LSTM is effective.The forecasting model is applicable not only to the short-term traffic flow forecast in the whole way,but also to the short-term traffic flow forecast of the single mode.