Study On The Factors Influencing Taxi Travel Demand Considering The Modifiable Areal Unit Problem

With the continuous expansion of the city,the travel distance of urban residents is increasing,resulting in taxi services playing a pivotal role among urban residents.Using taxi trajectory data or geographic location coordinate data to analyze the demand of taxi travel can improve the efficiency of taxi service and alleviate urban traffic congestion.The current transportation demand analysis is based on various types of zoning schemes,but demand analysis based on different zoning schemes will produce the modifiable areal unit problem(MAUP),which makes statistical model diagnosis and parameter estimation results unreliable.In order to study the impact of MAUP on taxi travel demand and the built environment factors,firstly,using 7 types of zoning schemes that are the official division of Census Block Group(CBG)and Census Tract(CT)and five scale grids division ranging from 400 to 2000 feet created by Arc GIS software as the statistical unit of data.Secondly,the taxi ridership per unit area in each unit during peak hours in Manhattan,New York City is used as travel demand to analyze the differences in travel space distribution under different zoning schemes and its spatial autocorrelation quantitatively.And then taking land use,population,social economy,transportation facilities and other related built environment data as influencing factors,analyze its multicollinearity and establish three models that are Spatial Lag,Spatial Error,and Spatial Durbin Model.The validity of the Spatial Durbin model is demonstrated through relevant statistical tests and the differences in related results are studied from the scale and zoning effects respectively.Finally,the estimated coefficients of the spatial Durbin model are decomposed into direct effects,spillover effects and total effects.The impact of MAUP on the three is studied and the results of each zoning schemes are considered comprehensively.The impact of influencing factors on the demand for taxi travel in the unit itself and surrounding units are analyzed.The research results indicate that: 1)The spatial autocorrelation of taxi travel demand and the collinearity of built environment variables will all be affected by the MAUP effect,and there are certain rules and patterns between the result differences and the zoning schemes.2)The MAUP effect does affect the choice of each zoning scheme data set between the spatial error model and the spatial lag model,but the conclusion that Spatial Durbin model is the best model will not be affected by MAUP.3)From the results of model prediction accuracy,it can be seen that the Spatial Durbin model based on the CBG zoning scheme fits the data best;the prediction accuracy results in the grid division method does not show that the smaller the scale is,the better result the model can be obtained.4)The estimated coefficients and decomposition effects of the Spatial Durbin model show that the significance and coefficients of the built environment variables will be affected by the MAUP effect,and the variation of coefficients has no discernable patten,but the variables in the smaller-scale zoning scheme are affected less by the zoning effect.The conclusions confirm that MAUP will affect the spatial data analysis results of taxi travel demand.Therefore,it is recommended that if transportation demand analysis is based on spatial units,the MAUP effect needs to be considered,and it also provides a reference factor for the division of traffic analysis zones.