Research On Subway Passenger Flow Prediction Based On Deep Learning

With the gradual complexity of the subway line structure,passenger flow is increasing.How to alleviate the congestion of rail lines is the core problem in intelligent transportation system.At present,a large number of studies are focused on road traffic,and it is difficult to provide accurate passenger flow information and decision-making basis for rail operation departments due to the lack of mining of the change mode of rail station flow and accurate cognition and evaluation of rail line flow.In view of the above problems,this paper uses the traffic pattern mining method based on spectrum periodic characteristics and statistical methods,machine learning,deep learning and other methods to study the rail traffic pattern and short-term traffic flow prediction.The main research contents of this paper are as follows.1.In view of the problem of traffic change pattern mining at rail stations,a traffic pattern mining method based on spectral periodicity is established.The K-Means clustering algorithm is modified by using the spectral characteristics to extract different modes of site traffic.Then the main function of the site area interest point(POI)distribution analysis mode is used to help understand the cluster mode.Finally,the effectiveness of the proposed method is verified by combining the real orbital data.The experimental results show that this model has better performance in time efficiency and clustering effect,and the clustering results of the model as a priori knowledge can improve the accuracy of flow prediction.2.Aiming at the problem of accurate evaluation of rail line flow,a short-term traffic flow prediction method based on CA-LSTM(The Cluster based Attention-LSTM Model)is established.On the basis of orbital flow periodicity and regional spatial connectivity,Attention-LSTM structure is established to learn orbital time series information.Based on the classification of rail transit modes based on spectral periodicity,the importance of traffic prediction is excavated by using Attention-LSTM structure according to historical dependence conditions.A short-term traffic flow prediction model based on CA-LSTM is proposed to predict rail transit traffic.The experimental results show that the prediction effect of CA-LSTM model is significantly improved.Through the above research,it can be seen that the traffic pattern mining method based on spectrum periodicity features has carried on the reasonable discrimination to the traffic pattern,and provides the high density and high demand track station distribution for the rail transit managers at different times;the short-term traffic flow prediction method based on CA-LSTM improves the prediction accuracy,which is of great significance to reasonably limit the access passenger flow and alleviate congestion.This study combines the research of flow mode and the prediction of passenger flow to effectively explore the development trend of rail traffic flow,and provide a basis for the decision-making of the establishment of rail stations and the optimization of rail train schedules,effectively alleviate travel congestion and improve the control of rail transit.