OD Prediction Of Urban Rail Transit Based On Spatial Data And Generalized Linear Bayes

In order to meet the demand of rapid and accurate OD prediction in urban rail transit network planning,the OD prediction method based on spatial data and generalized linear Bayes is proposed.Using spatial data to study the catchment area of stations and extract the influencing factor indicators of urban rail transit OD,establish a generalized linear model of OD volume and influencing factor indicators,and use Bayesian method to eatimate the parameters of generalized linear model to predict OD volume and its distribution.The main research contents include:(1)Based on the formation of urban rail transit OD and the process of rail transit passengers' travel,this paper analyzes the factors that affect the OD volume of urban rail transit,summarizes the characteristics of existing direct prediction methods of urban rail transit OD based on spatial data,and constructs a generalized linear Bayesian OD prediction idea and framework based on spatial data and multi-mode catchment area.(2)This paper analyzes the multi-mode catchment area of urban rail transit stations.Based on the data of investigation,trackof shared bicycle and the path planning API of Gaode map,a method for calculating the multi-mode connection time threshold is proposed based on the time characteristics of multi-mode connection and density clustering,and a model for determining the stations catchment area of walking and bicycle based on the connection time threshold is constructed.(3)Based on the spatial data of land use around the station and the catchment area of the station,the OD influencing factor indicators are constructed,and the correlation analysis and multicollinearity diagnosis are carried out.The random forest model is used for feature selection.Based on the OD influencing factor indicators after feature selection,the generalized linear model for OD prediction is constructed.(4)Bayes method is used to estimate the parameters of generalized linear model.According to Bayes theory,the posterior distribution of OD parameters under different likelihood functions,such as Poisson distribution,negative binomial distribution and gamma distribution,is derived respectively.The posterior parameter estimation results are obtained by sampling based on MCMC,the likelihood function is selected by Bayes information criterion,and OD is predicted according to the model selection results By analyzing the goodness of fit and error of OD prediction results,the results show that the OD prediction model established in this paper has better prediction effect and better prediction accuracy than the ordinary linear model.(5)Taking Beijing Metro Rail Transit Network in 2018 as an example,the OD volume prediction and result display are carried out to verify the prediction method in this paper.