Research On Taxi Seeking Strategy And Recommendation Model Based On Taxi GPS Data

In recent years,with the gradual integration and development of traditional cruise taxis and online taxis,the service model of drivers responding to order requests has been widely used.Online car-hailing has greatly improved the convenience and service quality of taxi travel,but the innovation of the service model has not significantly improved the income level of taxi drivers.It has become one of the unstable factors in the current taxi industry,and the driver’s income level is closely related to the driver’s efficiency in finding passengers.Therefore,exploring how to improve the efficiency of driver searching for passengers has become a topic worthy of research.Based on the driver’s perspective,this research aims to improve the efficiency of driver searching for passengers.(1)Firstly,it analyzes the data sources used by related researches,elaborates the disclosure of GPS data for taxis at home and abroad.By comparing the situation of taxis in New York City and domestic cities,it is clarified that the subsequent analysis methods and models constructed in this article have certain reference significance for domestic taxis.The rationality of the taxi operation data set,the preprocessing of the original data was completed after necessary explanations of the original data set;(2)Secondly,based on the New York City taxi operation data set in a random month in 2013,according to the driver’s income level and the day’s Ranking standard deviation,representative 419 low-income drivers,351 middleincome drivers,and 396 high-income drivers were extracted from all 31,950 drivers as research samples,and high-income drivers were extracted from two levels of time preference and regional preference;(3)Finally,based on the analysis of taxi operating costs and revenue composition,the study is based on the combination of the ordercarrying state and the adjacent empty-load state,and the order measurement is proposed.Based on the benefit index of the real revenue level of the order sequence,a driveroriented customer-seeking recommendation model was further constructed to predict the total benefit of some future orders for drivers,instead of being limited to the onetime benefit of the next order.Considering whether the driver adopts the current recommended order sequence,the dynamic update method of the model is further designed,and the effectiveness of the model is verified by using the New York City taxi operation data set from January to July 2013 as an example.This study first explored the unique customer-seeking strategies of high-income drivers based on real taxi operation data in a random month in New York City in 2013.The results show that high-income drivers in New York City mainly work during the day.Compared with other drivers,high-income drivers have the main operating area is also more closely fixed in the central area of Manhattan.At the same time,high-income drivers will deliberately control the area where the pick-up and drop-off points are located to avoid difficulty in receiving the next order.This also breaks the traditional impression that it is more profitable for taxi drivers to take long-distance orders,and that taking orders at night can avoid traffic jams and improve the efficiency of taking orders;secondly,a driver-oriented customer search recommendation model was constructed,and the model was carried out using real taxi operating data in New York City from January to July 2013 as an example.Example verification shows that the model can effectively help drivers increase their income.