Research On Short-term Inbound Passenger Flows Prediction Of Urban Rail Transit Based On Deep Learning

With the continuous development of our country’s social economy and urbanization,the construction of urban transportation infrastructure is lagging behind,and the supply capacity is insufficient,which makes it difficult to meet the growth of residents’ transportation demand,resulting in more and more serious urban traffic congestion problems.With its advantages of large capacity,punctuality and high efficiency,rail transit is one of the most effective measures to alleviate urban traffic congestion.With the continuous construction of rail transit and the gradual formation of the rail network,the proportion of rail travel has increased year by year,which has brought great challenges to the operation and organization of rail transit and put forward higher requirements for the operation and service of rail transit.The inbound passenger flow of rail transit,especially the short-term inbound passenger flow,is the basis for the formulation of rail transit operation organization measures,and is crucial to improving the management level and service level of rail transit operation.In this paper,based on rail transit AFC swipe card data,it is important to construct a short-time inbound passenger flow prediction model using deep learning modeling methods to study short-time inbound passenger flow prediction.Firstly,the rail transit AFC swipe card data is pre-processed,data processing and passenger flow extraction is achieved using programming languages such as Python in this paper.Secondly,the factors influencing rail transit passenger flow are analysed,and the characteristics of rail transit passenger flow are analysed from the time and space dimensions,showing fluctuation and periodicity in time and distribution differences and imbalances in space;based on the extracted passenger flow data in and out of rail transit stations,rail transit stations are classified into six types: residence-oriented,employment-oriented,dislocation-biased residential,dislocation-biased employment,employment-residential dislocation,and comprehensive,and the passenger flow patterns of different types of stations are identified.Then,considering the influence of the prior sequence passenger flow and inter-station passenger flow on the prediction of short-time inbound passenger flow,a multi-temporal feature LSTM model and a CNN-LSTM-ATT model are constructed to predict short-time inbound passenger flow using deep learning modeling methods,and the whale optimization algorithm(WOA)is used for parameter optimization.Finally,the construction of a multi-temporal feature LSTM model and a CNN-LSTM-ATT model are applied to predict short-time inbound passenger flow at stations with different passenger flow patterns in the Chongqing rail network,and common non-linear models such as BP,SVM,CNN and LSTM are selected for comparison.The results show that the multi-temporal feature LSTM model constructed in this paper is 6.63%～36.94% less than the baseline model in terms of root mean square error(RMSE)and 7.43%～31.91% less than the baseline model in terms of mean absolute error(MAE);the CNN-LSTM-ATT model constructed in this paper is5.34%～36.05% less than the baseline model in terms of root mean square error(RMSE)and 1.33%～38.91% less than the baseline model in terms of mean absolute error(MAE).It can be seen that the prediction effect of the short-time inbound passenger flow prediction model constructed in this paper has been improved in various degrees compared with the baseline model,with higher prediction accuracy and better applicability.In this paper,considering the influence of prior sequence passenger flow and inter-station passenger flow on short-time inbound passenger flow prediction,the multi-temporal feature LSTM model and CNN-LSTM-ATT model are constructed by using deep learning modeling methods,which can extract the temporal features of inbound passenger flow and spatial features of inter-station passenger flow of rail transit,grasp the change pattern of rail transit passenger flow,improve the prediction accuracy of short-time inbound passenger flow of rail transit,enrich the theory of short-time passenger flow prediction of rail transit,and provide data support for the operation and management related decision of rail transit department.