Research On Short-Time Passenger Flow Prediction Of Urban Rail Transit Line Network Based On Spatio-Temporal Characteristics

With the increase of urbanization,the passenger flow of urban rail transport is rising and the scale of the network is expanding at a high speed.The spatial and temporal characteristics of urban rail traffic are becoming increasingly complex with the expansion of the network scale,and the mismatch between traffic demand and supply in time and space is becoming more and more serious,posing a huge challenge to operation management.The development of big data processing and deep learning technology has laid the foundation for solving the problem.The OD passenger flow in the short-term line network includes both the temporal and spatial distribution characteristics of passenger flow information,which can clearly and dynamically characterise the travel demand and changes of passengers.In view of the scarcity of research on short-duration OD passenger flows on line networks,the fragmented spatial and temporal characteristics of passenger flows,and the low prediction accuracy,the thesis uses AFC data to analyze the spatial and temporal characteristics of passenger flows and establishes a prediction model to forecast short-duration OD passenger flows by combining the characteristic laws.The research content of the thesis is as follows:(1)The AFC data of Hangzhou city rail transit in January 2019 were pre-processed and the station adjacencies of the rail line network were analyzed simultaneously.The basic characteristics of urban rail passenger flow are analyzed according to the progressive hierarchy of stations,lines and line networks,while the data format is transformed and the entry AFC swipe data is converted into OD matrix data,replacing the traditional raster image data processing method.(2)From the perspective of temporal characteristics of the line network passenger flow analysis flow temporal regularity characteristics,the use of cohesive hierarchical clustering method for the flow of daily data type calibration,using Pearson correlation coefficient calculation method to analyze the correlation between the predicted time period flow and the front input time period flow and determine the best front correlation time period,based on the introduction of the best front correlation time period to analyze the data time granularity.The hidden spatial attributes behind the temporal types are explored from the perspective of spatial characteristics,peak types are classified and site types are determined according to the distribution of temporal types.The sparsity of the data is reduced using OD attractiveness and the types of OD pairs are calibrated.(3)A short-time OD passenger flow prediction model for urban rail line networks based on spatio-temporal characteristics and incorporating an attention mechanism is proposed.A two-threaded convolution operation is proposed to thicken the data by using the attention mechanism to calibrate the weights of the input OD data for multiple time periods.The model considers the previously proposed optimal antecedent correlation time period,time granularity,and forecasts two sets of data according to two types of daily data for intra-week and weekend respectively.The model prediction result metrics are compared and analyzed with those of other prediction models to evaluate the model performance and validate the value of considering spatial features and fusing attention mechanisms to improve the accuracy of passenger flow prediction.The results show that the introduction of the attention mechanism allows the model to accurately assign feature weights,resulting in a 43.58% reduction in RMSE on the average intra-weekly data and 41.41% reduction in RMSE on the weekend data at all time granularities.The predicted OD image brightness peaks are accurate and clear,indicating that analyzing the spatio-temporal features and applying them to the OD matrix can effectively reduce the overfitting phenomenon that complicates the processing model.Dynamic and accurate prediction results are important for rail operators to rationalize their operational plans to achieve efficiency and reduce the cost of passenger travel time.