Research On Multi-space Scale Passenger Flow Prediction For Chongqing Rail Transit Based On Deep Learning

With the development of the urban economy,the urbanization process has been accelerating,and the prosperity of the urban economy has promoted the growth of the urban population.The city's inherent traffic capacity is no longer able to meet the growing travel needs of urban residents.In recent years,in order to meet the needs of residents' travel,the scale of urban rail transit network has been continuously expanded,and the scientific and rational operation of the network has become the top priority of the development of the transportation industry.Therefore,this thesis establishes and trains a multi-space-scale passenger flow prediction model,which is used to predict the passenger flow of the inbound and outbound passenger flow,the site OD passenger flow,and the regional OD passenger flow,and the traffic flow of the three different spatial scales.Data transmission support and regional line network planning provide data support,thereby realizing the rational operation of the entire network.The main work of this thesis is as follows:(1)Analyzed the development of urban rail transit,expounded the significance of multi-space scale passenger flow forecasting,and studied the current passenger flow forecasting theory and technology.(2)Selecting the historical real passenger flow data of Chongqing rail transit as the research object,and analyzing the space-time characteristics of passenger flow.(3)Four neural network models with different spatial scales are constructed,which are the LSTM model of passenger flow forecasting based on space-time influence flow matrix,the passenger flow forecasting model based on residual neural network,and the historical synchronization and sequential time series.The site OD passenger flow prediction LSTM model and the passenger flow prediction LSTM model based on regional division and passenger travel mode.(4)In order to construct a passenger flow prediction model based on regional division and travel mode,the division of the region is first carried out,and then the travel mode is identified.(5)Based on the orbital passenger flow data between 2017 and 2018,this thesis preprocesses the data,extracts the journey time,sequence time series,historical synchronic sequence and spatiotemporal influence matrix,and substitutes the established four different passenger flows.The prediction model was tested.The model prediction results at different spatial scales are analyzed and compared with commonly used models such as SVR and RNN.The experimental results show that the passenger flow prediction of each scale is better than the common passenger flow prediction algorithms such as SVR and RNN.The RMSE and MRE errors are smaller than the common passenger flow prediction algorithm.And by comparing the passenger flow prediction results of different scales,it is found that with the expansion of the scale,the passengers travel more regularly,and the error of the prediction results is smaller.