Modeling Location Choice of Taxi Drivers for Passenger Pickup Using GPS Data

Recently, the traditional taxi industry has been struggling to keep its market share, especially with the emergence of new transport network companies (e.g., Uber). One of the problems with traditional taxi services is the difficulty of matching the taxi demand to its supply when there is no phone-booking or other reservation system. From that perspective, the taxi driver?s experience is important in reaching the next passenger. A taxi driver with limited experience may not know the high-demand locations and times of taxi stands or street sections to visit after dropping off a passenger. This causes a large number of vacant taxi drivers to regularly cruise the roads to search for a passenger, contributing to congestion, pollution, and resource waste. We formulate the problem of a taxi driver?s next passenger pickup location as a destination choice problem. Vacant taxi trips between drop-off and pickup points are extracted from GPS records obtained from a taxi operator in Lisbon, Portugal, to understand the travel behavior of vacant taxi drivers. We have estimated destination choice models with a multinomial logit and a nested logit structure. It was found that passenger demand at the pickup area, hotspot locations, service location preference, and major transport hubs positively influence a taxi driver?s next choice of passenger pickup location. Results of this study provide insight regarding the factors that explain a taxi driver?s probability to choose a certain zone within a set of passenger pickup zones, contributing to a better understanding of taxi driver travel behavior.