Research on nonlinear relationship between subway built environment and travel distance of stations based on XGBOOST-SHAP
Compared to traditional analysis of passenger flow characteristics,the consideration of the average travel distance of metro stations leads to a more refined understanding of the passenger flow dynamics of networks.This study focused on the metro system of Xi'an City to explore the complex relationship between multiple built environment factors and the station-level average travel distance.Eleven built environment indicator systems,including land use,point of interest features,surrounding transportation infrastructure,and station attributes,were first developed.An interpretable machine learning model based on Extreme Gradient Boosting with SHAP attribution analysis framework(XGBOOST-SHAP)was established to reveal the nonlinear relationship between these factors.Additionally,the advantages of the XGBOOST model in regression fitting were verified by the comparison with Gradient Boosting Decision Trees(GBDT)and Ordinary Least Squares(OLS).The results show that the XGBOOST model achieves an R-squared value of 0.75,with a mean absolute error(MAE)of 0.95 and mean squared error(MSE)of 1.36,outperforming the GBDT and OLS models in terms of fitting performance.A clear circular distribution pattern can be found with the spatial heterogeneity of average travel distance.SHAP attribution analysis reveals that apart from the distance to the city center feature,other features such as road network density,land use mix,the number of bus routes,and residential count also contribute significantly to the travel distance.The influence of POI Shannon entropy index and food service points on average travel distance does not show clear positive or negative feedback.Other indicators demonstrate a combined positive and negative feedback mechanism on average travel distance.The research results,which are beneficial for transportation demand analysis,route capacity optimization,and operational effectiveness evaluation,can effectively improve the convenience of metro transportation,satisfy the needs of different regions,and enhance the efficiency and sustainability of the entire metro system.
metro stationsbuilt environmenttravel distanceXGBOOST modelSHAP attribution analysisnonlinear relationship
国家科技期刊平台
NSTL
万方数据
