Research On Influencing Factors Of High-income Taxi Drivers' Income Based On GPS Data

The income of the taxi driver is closely related to the following three aspects: no-load distance,passenger distance and route selection.The experienced drivers can better balance the influence of these three aspects to the income than the newly-entered drivers.Our work is aimed to identify high-income drivers from taxi GPS time-series trajectory data and extract the characteristics of high-income drivers in these three aspects.First of all,in order to identify high-income drivers from taxi GPS data,the GPS track data is divided to two kind of Trip track points sets based on the Original-Destination(OD)flag.They are passenger Trip track points and no-load passenger Trip track points respectively.However,due to the offset phenomenon of GPS data and the low sampling frequency,the Trip track points sets do not fully indicate the corresponding actual path.To this end,the road network data structure map is established by using the urban road feature points provided by OpenStreetMap,and the Trip track point set of the GPS data is converted into the actual Trip path to estimate the taxi driver income.Then,the K-Means clustering algorithm is used to classify taxi driver's income into three levels: high-income,medium-income and low-income.The characteristics affecting the income of high-income drivers are firstly analyzed by Logistic regression model.The correlation between different income levels and the no-load distance,the passenger distance and the passenger taxi speed are obtained respectively.These three factors are used to judge the driver's operation processes,and driver's operation processes is regarded as important thresholds.Then using these as clue to screen out highincome driver's no-load Trip path set and passenger Trip path set.Next,Dividing a day into different time segments,classifying the path set into different time segments according to the path start point timestamp,and using the DBSCAN algorithm to set the starting point of each path in the no-load Trip path set and the passenger Trip path set in the same time period.Clustering is performed to obtain an unloading point that can search for passengers at a shorter distance under a specific time-space conditions,and a potential high-value passenger point.The high-income driver Trip path set filtered by the speed evaluation threshold can determine the unobstructed path of the city under specific time and space conditions.The research result shows that characteristics of high-income drivers can identify shorter no-load distance,high-value passenger point identification and unobstructed path selection after passengers in a specific time and space.Improving the income level has instruction and application value,and it can be used as a reference for taxi management department and government to issue corresponding management policies.