Passenger Route Choice In Urban Rail Transit Based On Spatio-temporal Network And AFC Data

With the intensive opening of new urban rail transit lines and the continuous advancement of networked operations in many metropolises of China,the number of routes available to passengers between OD pairs and the proportion of transfer flow in the entire network are increasing.Under the seamless transfer operation,grasping the route selection of passengers and the spatio-temporal distribution of passenger flow is critical for balancing passenger flow distribution and improving operational efficiency.This thesis focuses on the problem of passenger route choice between multi-path "fuzzy transfer" OD in urban rail transit.Based on the spatio-temporal network to describe the effective path under a specific spatio-temporal prism,passenger trajectory is deduced through the analysis analysis of AFC data.The route choice behavior is based on free time to estimate an individual’s decision-making.Using supervised learning methods,a decision tree model is constructed,which provides a model and data support for an accurate,efficient and rapid grasp of passenger flow.This thesis mainly researches on the following four aspects:(1)Demonstrate the time and space limitations of passenger travel.Based on the theory of time and space network and the expansion of train timetable information,the urban rail transit spatio-temporal network from the perspective of passengers is built,and a three-layer k(k≤3)short path depth-first algorithm realizes the matching of passenger effective path set generation and spatio-temporal travel trajectory deduction.(2)Through the AFC data extraction,transformation and cleaning,establish urban rail transit operation data warehouse.Mining the sample data of reference passengers,the tolerance intervals of entry time,exit time and transfer time in the corresponding direction under 95% confidence are obtained,which clears the obstacles for the drawing of entry arc,transfer arc and exit arc.(3)From the view of passengers,analyze the free time that is critical to restricte passenger route choice,which is based on the spatio-temporal network and effective spatio-temporal routes.Through the destination backstepping and the lower limit of tolerance interval,the route choice behavior between multi-path ODs could be efficiently estimated.The innovation summarizes the structure and characteristics of 6 basic units of transfer micro-network,and proposes the concept of "fuzzy transfer".(4)Summarize the main attributes which affect the passenger’s route selection decision.Select the typical "fuzzy transfer" OD pair data of Chengdu Metro,and quantitatively analyze the impact of each attribute on route selection.Using this as a training data set,a decision tree for rapid classification and prediction of passenger path selection was generated.It was found that the increase or decrease of the number of transfers of each path was the most important factor affecting the passenger’s selection of the shortest path in theory.Through the performance evaluation and inspection,the efficiency of the decision tree model is verified.