Understanding taxi driver’s cruising behavior with ZIP model

Abstract

The taxi drivers’ cruising pattern was learned using GPS trajectory data collected in Shenzhen, China. By employing zero-inflated Poisson model, the impacts of land use and previous pick-up experience on cruising decision were measured. The cruising strategies of different types of drivers as well as the top one driver were examined. The results indicate that both land use and previous pick-up experience affect travel behavior with the former’s influence (7.07×10−4 measured by one of the coefficients in zero-inflated Poisson model) being greater than the latter’s (4.58×10−t) in general, but the comparison also varies across the types of drivers. Besides, taxi drivers’ day-to-day learning feature is also proved by the results. According to comparison of the cruising behavior of the most efficient and inefficient driver, an efficient cruising strategy was proposed, that is, obeying the distribution of land use in choice of cruising area, while learning from pick-up experience in selection of detailed cruising location. By learning taxi drivers’ cruising pattern, the development of measures of regulating travel behaviors is facilitated, important factor for traffic organization and planning is identified, and an efficient cruising strategy for taxi drivers is provided.