Analyzing Taxi Customer-search Behavior Using Copula-based Joint Model

As an important part of urban traffic,taxi can provide convenient and fast service for residents to travel.However,due to the mismatch between supply and demand in time and space,a large number of empty taxis cruise on the road,which exacerbates traffic and environmental problems.Therefore,the study of taxi operation behavior is an important issue in traffic management and taxi scheduling.Based on the GPS trajectory data of taxis,this thesis studies taxi customer-search behavior.The main research results include:First,for the complex customer-search behavior,the traffic periods are divided according to the distribution of taxi average speed.DBSCAN clustering algorithm is used to conduct clustering analysis on the pick-up and drop-off points of passengers.On the basis of OD division,the frequency of taxi paths was counted to generate candidate routes.this thesis chose the destination customers generation rate,destination distance,route customers generation rate,route length,the accumulative intersection delay,route travel time and path size seven variables from the perspective of costs and benefits,and seven variables are calculated based on taxi trajectory data.The influences of various variables on taxi drivers’ destination choice and route choice are analyzed.Second,as for the limitation of separating taxi drivers’ regional choice and route choice,a copula-based joint model is developed to analyze destinations choice and routes choice behavior in the customersearching process.A multinomial logit model(MNL)is used to analyze the destinations selection behavior,and a path size logit model(PSL)is used to explore the routes choice behavior.Then we use copulas function to connect the two models,and apply the Bayesian information criterion calculation of goodness-of-fit.Finally,the model parameters are solved by the maximum likelihood method,and the decision-making mechanism in taxi operation is analyzed.Third,for the taxi customer-search behavior in different urban land use areas,this paper divides the research area into 144 traffic zones and identifies the land use types of traffic zones.Then we selected six indicators to measure the economic and social differences of traffic zones with different land use types.Finally,multinomial logit model is used to analyze taxi selection behavior,and the results of parameter calibration are discussed in depth.There are 30 figures,13 tables and 80 references.