Recommendation Model Study Based On Taxi Trajectory Data Mining

With the rapid development of the taxi industry, taxis have made up for the deficiency of public transport and private transport, became one of the necessary traffic tools in the daily travel, and played an important role in the transportation. At present, the traffic in Beijing is gradually deteriorating. On the one hand, "difficult to find an empty taxi" in rush hours becomes a general phenomenon of transportation. And at the same time, many taxis travel on the road without any passengers,which is called“empty traveling”. The “empty traveling” brings a lot of troubles,such as it can improve the operation cost of the taxi driver, reduce revenue, bring extra burden to the existing traffic conditions, and aggravate the environment pollution and energy waste problem. Based on the above reasons, how to effectively improve the balance between taxi supply and passenger demand becomes a problem urgently to be solved.In recent years, for the sake of safety and convenient, many big cities such as Shanghai, Sazhen, Beijing have equipped with the GPS equipment on the cab. These GPS devices will send massage to the taxi management center at a certain frequency.The message consist the latitude and longitude of the vehicles, time, and passenger state, etc. These data accumulations and generates a lot of taxi GPS trajectory data. Taxi GPS trajectory data includes two aspects of content, on the one hand is the taxi passenger travel information, such as when passengers get on the bus,get off, the origin and destination of travel, etc.,which can be used to study People’s daily travel behavior. On the other hand, the data includes the taxi drivers’ habits and behaviors, such as where to find the driving route, the shortest route, which can be used to study the shortest path of city road and traffic congestion.Based on the research of the above two aspects, in this paper, the author puts forward a recommended model for passengers to find the best drop off point and a recommended model for the taxi driver to find compatible passengers. Including two aspects of content: on the one hand, by a large number of GPS trajectory data mining,this paper finds taxi driver behavior characteristics, recommends the passengers walking to a latest waiting point to get a car so as to effectively reduce the passenger waiting time. These sites called taxi stands. On the other hand, by improving the traditional CKNN algorithm, the author raises a new CKNN algorithm to find the candidates for the taxi drivers. In this paper, the author puts forward thereare two factors which can affect taxi drivers’ selection. The two factors are the passenger destination’s popularity and driver preferences relevance.By using BP neural network algorithm,the author founds a recommend model to help the taxi drivers’ to make their decision by recommending passengers.This paper’s significant contributes are as bellowed:This article embarks from the taxi GPS trajectory data and finds higher profits taxi drivers by clustering algorithm mining. Through studying from these drivers,who parking to wait for passengers in non-rush hours, this paper puts forward a new algorithm to find taxi stands.After creates a recommend model, the author offers latest taxi stands’ position to the passengers who want to take a taxi.In this article,the original CKNN algorithm has been improved to rapidly find the nearest passengers around the taxis.Through the taxi GPS trajectory data mining, the two influent factors of taxi drivers’ selection has been found and calculated.By using BP neural network algorithm, the author sets up his his passenger recommend model to recommend the best candidate for the taxi drivers.