Traffic Flow Prediction LSTM Attention Mechanism Neural Network Research On LSTM Short-term Traffic Flow Prediction

In recent years,with the rapid development of China’s economy,the urbanization process is also accelerating,and the per capita vehicle ownership is also increasing.The problem of traffic congestion broke out one after another,seriously affecting the normal life of urban residents.Therefore,reasonable road planning and traffic facility layout,as well as the adjustment and real-time management of traffic flow have become the main tasks of road traffic management.The prediction of short-term periodic traffic flow can not only provide the theoretical basis of traffic diversion,but also help people make more travel decisions,alleviate traffic congestion,reduce carbon emissions and improve traffic operation efficiency.Artificial neural network is a new research hotspot in recent years.It is an information processing concept that imitates the operation mode of neurons in the human brain.Through the establishment of different models,it processes the operation of neurons for different information types.Like the human brain,neural network has the functions of memory and forgetting.At present,the research on neural network has made rapid progress,and the artificial neural network system has been mature and applied to various aspects.In this paper,artificial neural network is applied to the prediction of Expressway short-term traffic flow.After analysis and research,attention mechanism is introduced to study the accuracy and efficiency of prediction.The research contents of this paper contain: In order to enhance the applicability of the data,the data obtained by WIM dynamic weighing system is analyzed and processed,and a convenient and efficient method for processing this data is given.The system detection data is transformed into traffic flow data in a short time,and the data is integrated to make the data better suitable for neural network system.(1)In order to enhance the applicability of the data,the data obtained by WIM dynamic weighing system is analyzed and processed,and a convenient and efficient method for processing this data is given.The system detection data is transformed into traffic flow data in a short time,and the data is integrated to make the data better suitable for neural network system.(2)This paper introduces the traffic flow theory,analyzes the influencing factors of traffic flow.At the same time,this paper analyzes the characteristics and internal structure of several existing neural network models,analyzes the modeling rules of traffic flow prediction,and puts forward an LSTM neural network model which is most suitable for traffic flow prediction in a short time.(3)After introducing the attention mechanism and analyzing the data characteristics,classify the attention mechanism,select the structure type of attention mechanism suitable for traffic flow data,combine the attention mechanism with LSTM artificial neural network model,analyze the operation characteristics of neural network,and use dropout to solve the over fitting problem of network,This paper expounds the theory and internal structure characteristics of LSTM model under the action of attention mechanism,as well as the specific training methods of network model.(4)The proposed model is verified by data experiments.Firstly,the different types of the internal structure of the neural network are optimized,and then the effectiveness of the attention mechanism is verified.A variety of evaluation mechanisms are used to compare the proposed neural network model with other traditional neural network models.After comparison,the model proposed in this paper has significantly higher prediction accuracy and running speed.