Current Awardee

Improving the Efficiency of Taxi Systems through Real-time Seek Time Prediction

Daniel Work

College: Engineering
Award year: 2014-2015

The goal of this project is to assess if the seek time for a taxi is predictable, and if so, to also learn the predictors. The seek time for a taxi is the time between when the previous passenger was dropped off and when the next passenger is picked up. High driver income correlates with low seek times between trips. Obviously all taxi drivers would prefer to reduce their seek time, however it is difficult for the drivers to predict where to go next because the seek time is a function of the passenger demand, and a function of competing taxis that may pick up the passenger first. To accurately predict the seek time throughout the city, taxi drivers must be able to predict both the changes in the demand for trips, and the reactions of the other taxi drivers (i.e. the supply). This work will examine a dataset of more than 700 million taxi trips in New York City to predict seek times, ultimately enabling predictive taxi information services to improve the overall system efficiency. Such a service would benefit both the passengers of the taxi system with reduced wait times, as well as the drivers through higher income.