| With the rapid development of domestic economy,the operation mode of urban rail transit has changed from the era of single line to the era of network.The travel of passengers in urban rail transit network is becoming more and more diversified,and passengers will have multiple paths to choose from for each trip;Moreover,the heterogeneity of passengers has become increasingly prominent,which has brought great difficulties to the passenger flow sorting of urban rail transit.A clear and reasonable description of the physical path selection behavior of passengers in the operating road network is conducive to grasping the distribution law and characteristics of passengers in the road network,and it provides a scientific theoretical basis for the relevant operational departments to formulate reasonable operational strategies.At the same time,with the popularity of automatic fare collection system and all-in-one card,the passenger flow data is accumulating at an amazing rate,and the passenger flow in and out of the station data are easier to obtain,and it provides practical and reliable data support for the study of passenger travel behavior in urban rail transit network.This paper uses the passenger flow data generated by AFC(Automatic Fare Collection System),starting with the mining of passenger travel characteristics,classifies the passengers from the perspective of travel law,and uses the improved logit route selection model to estimate and predict the passengers’ travel paths on the road network.Firstly,based on the real travel records of passengers,the characteristics of passengers are extracted and analyzed by means of data mining.The characteristics of travel intensity,time dimension,spatial latitude and card type are extracted from the one month passenger travel AFC data information.Based on the RFM index,the previous research is innovated and supplemented to construct an objective passenger classification index.The sample passengers within one day are classified by SPSS software using two-step clustering algorithm,and the optimal clustering number is selected by BIC criterion to divide the passengers into five categories.Then,according to the various index characteristics of all kinds of passengers,analyze and summarize the travel behavior law of passengers,so as to preliminarily identify their socio-economic attributes.Secondly,the route selection model of urban rail transit is established based on MNL model(Multinominal Logit Model).On the basis of considering the influence of road network familiarity,travel time,transfer complexity and other factors,this paper constructs the generalized cost function,improves the logit model and builds the algorithm framework on the basis of using the depth first search algorithm of graph to search the effective path of the whole network.In terms of parameter calibration,the parameter calibration method based on Bayesian inference is used for parameter calibration,and then the MCMC algorithm is used to estimate the parameter values in combination with the characteristic attributes of various passengers and AFC data information,so as to obtain the parameter set of various types of passengers.Finally,Taking Zhengzhou rail transit network as an example,the above algorithm and model are used to match the path of each passenger,and the selection probability of each path is integrated to verify the feasibility and applicability of the model. |
