Research On Short-term Traffic Flow Forecast Technology Based On ES-AFS-LSTM

With the progress of modern science and technology,various means of transportation bring unprecedented speed and convenience to people’s travel.Meanwhile,the number and frequency of transportation has increased sharply as people’s dependence.The limited urban resources cannot meet the surging demand for traffic,causing serious traffic congestion and hindering the further development of the city.In view of the above problems,this paper uses the intelligent transportation technology to make effective short-term traffic flow prediction based on the pre-processed traffic flow data,so as to provide real-time traffic information for travelers and alleviate the inconvenience caused by traffic congestion.Firstly,in order to solve the quality problems existing in the original data,according to the analysis of traffic flow characteristics,the abnormal data in the original data were detected and eliminated.Then,according to the spatio-temporal characteristics of traffic data itself,a missing data repair method based on the improved historical trend method is proposed.By taking advantage of the similarity within the data collected by the near neighbor monitor,the eliminated data and the original missing data are repaired by the proposed weighted strategy.Finally,data normalization is carried out.Secondly,in order to improve the accuracy of short-term traffic flow prediction and make up for the shortcomings of single forecasting model,a combined model of Exponential Smoothing(ES)and optimized Long Short-term Memory(LSTM)is proposed.Artificial Fish-Swarm(AFS)algorithm is introduced to solve the shortcoming of traditional LSTM which is easy to fall into local optimum when using back propagation algorithm.Considering the basic characteristics of short-term traffic flow such as nonlinearity,randomness and uncertainty,a short-term traffic flow prediction model based on ES-AFS-LSTM is constructed.To verify the prediction effect of the model,relevant experiments are carried out on the measured data,and the prediction results of various models are compared and analyzed,which demonstrate that ES-AFS-LSTM attains superior performance compared with state-of-the-art methods.In this paper,the short-time traffic flow prediction technology is studied based on the measured traffic flow data,aiming to provide a solid theoretical basis for the development and application of intelligent transportation system,which has a very important significance and application value for alleviating traffic congestion and promoting urban development.