Research On The Behavior Of Customer-search Based On Taxi GPS Trajectory Data

Taxi as an important means of urban public transportation.On the one hand,it brings great convenience to residents’ travel with the flexible operation characteristics.On the other hand,the phenomenon of unoccupied taxi caused by the imbalance of supply and demand aggravates the traffic congestion and energy consumption in the city,which has caused trouble to the operation and management of taxis.Therefore,through the in-depth study of urban unoccupied taxi customer seeking behavior,taxi managers can accurately grasp the operation rules of urban taxi,so as to improve the efficiency of taxi operation and management,and alleviate the traffic problems caused by unoccupied taxi to urban roads.This article analyzes the behavior of customer searching for a taxi based on GPS data from Chengdu.The specific content includes three main aspects:(1)A track cleaning method was established to process the original taxi GPS trajectory data in Chengdu,select reasonable indicators to analyze the taxi operation characteristics,and propose a DBSCAN spatial clustering algorithm based on Euclidean distance to further explore the temporal and spatial distribution characteristics of pick-up and drop-off hot spots.(2)A dual simulation framework from region to route is proposed to analyze the customer seeking behavior of unoccupied taxi.On the regional scale,the road network is divided into grids of the same size,and each grid is regarded as a region.The region selection based on Logit model is established to analyze the search behavior of unoccupied taxis in the region and across regions.On the route scale,in order to solve the problem of route duplication in the process of customer seeking behavior simulation,the general characteristics of each candidate route and the benefit characteristics of the actual traffic scene are comprehensively considered.In order to solve this problem,this paper construct a PS-Logit model to simulate the route selection behavior of unoccupied taxi,selecting four influencing factors:unoccupied travel time,travel distance,passenger generation probability and traffic congestion index.(3)The dual simulation framework is verified and analyzed by using GPS data of Chengdu unoccupied taxi.The results show that 30.25%of the grid unoccupied taxis in the area have failed to find passengers successfully,and about 80%of the number of unoccupied taxis in the grid area has less than 40%success probability;in the cross region search process,when the number of cruise grids reaches 4,the cumulative probability of customer search success reaches more than 85%,and when the number of grids exceeds 8,the success probability of customer search is greater than 97%.Through the comparative analysis of the two types of characteristics that affect the route choice of unoccupied taxis,it is found that compared with the general characteristics of the route,the model performance is improved after the introduction of the route’s benefit characteristics,indicating that the route’s benefit characteristics are more important for the driver’s passenger routing influences.In addition,the calibration results of the MNL model and the PS-Logit model under different feature combinations are compared and analyzed,and it is found that the PS-Logit model has a better fit.