A Traffic Flow Prediction Model Based On Multi-dimensional Data And Long Short-term Memory Network

With the improvement of urbanization and the rapid development of economy,the demand for transportation is increasing rapidly,and the problem of traffic congestion is becoming more and more serious globally,which has become a global traffic problem.In order to relieve the pressure of urban traffic,many cities around the world have adopted intelligent traffic systems to direct traffic and solve urban congestion.Intelligent transportation system(ITS)can effectively alleviate traffic congestion under the condition of existing traffic network.Real-time and high-precision short-term traffic flow forecasting is one of the key technologies of ITS,and also the premise and foundation of ITS.Improving the accuracy of short-term traffic flow prediction has an important milestone significance and practical value for enriching the theoretical system of urban road traffic and enhancing the overall application of intelligent transportation,which can play a guiding role for traffic managers to effectively guide traffic,and provide decision-making basis for traffic travelers to grasp real-time,accurate traffic status and future trend information.In recent years,scholars at home and abroad have conducted a lot of research on short-term traffic flow prediction and proposed many innovative short-term traffic flow prediction methods.However,the most research is only focus on the optimization of models and algorithms,ignoring the dynamic interference of related factors in the traffic system,which makes it difficult to improve the prediction accuracy of the models under special conditions.Therefore,this paper proposes a deep-learning short-time traffic flow prediction model MDLSTM that combines the relevant influencing factors of traffic system and considers the temporal and spatial characteristics of the traffic flow to solve the problem of low prediction accuracy caused by the influence of traffic complex system-related coupling factors,to achieve real-time accurate short-time traffic flow prediction and provide a basis for intelligent transportation.Firstly,based on the analysis of the temporal and spatial characteristics of traffic flow,the model analyzes the traffic flow characteristics of urban traffic system under the influence of workdays,holidays,weather and other related factors,quantifies the related factors,and constructs a multi-dimensional state vector combined with road traffic flow.Secondly,to solve the problem that traditional Recurrent Neural Network can’t model the long-term dependent information in the traffic flow sequence,this paper use LSTM in deep learning as the basic structure of the model,and establishes the short-term traffic flow prediction model in combination with the multi-dimensional state vector.In order to verify the performance of the MDLSTM model,a large number of experiments were carried out in this paper using the real-time traffic flow data provided by the traffic research data laboratory of the United States.The results show that this method can avoid the problem of low prediction accuracy caused by the impact of various traffic system elements,and compared with the LSTM and BP neural network models based on historical traffic data,MDLSTM is more accurate and can reflect the future trend of traffic flow more accurately,has some practical value.Thirdly,considering the correlation and delay of the traffic flow of the upstream and downstream roads,and under different conditions,such as morning and evening peak and flat peak time,working days and weekends,the delay time will change.On the basis of the above research,in this paper,the best delay time series is determined by calculating the cross-correlation coefficient of upstream and downstream traffic flow in different states,and the next time optimal upstream and downstream delay time is predicted by combining multi-factor data.Then reconstruct the upstream traffic flow.Finally,the reconstructed upstream traffic flow is used as the supplementary flow for the prediction of the downstream road,so as to effectively guide the prediction of the short-term traffic flow of the downstream road.Experiments show that this method can effectively improve the accuracy of short-term traffic flow prediction.