PrivateHunt

Recently, for-hire vehicle services (FHV, e.g., Uber and Lyft) have become essential to people's daily transportation. Similar to taxis, how to effectively dispatch these FHV based on demand and supply is important for both FHV passengers and drivers. Based on real-world multi-source data, we identify two new challenges for FHV dispatching: (i) diverse demand: FHV passengers are a mix of passengers previously using taxis, buses, subways, or private vehicles; (ii) uncertain supply: FHV drivers join and leave the FHV system with spatiotemporal dynamics. As a result, the state-of-the-art taxi dispatching techniques cannot be applied to FHV systems. In this paper, we design the first FHV dispatching system PrivateHunt based on extremely large-scale urban transportation data from New York City and Shenzhen in China. In particular, we present (i) a passenger demand model based on taxi, bus, subway, and private vehicle data; (ii) a driver supply model based on small-scale FHV data; (iii) a dispatching technique for FHV vehicles based on proposed demand/supply models to reduce idle driving time. We implement PrivateHunt based on 14 thousand taxis, 13 thousand buses, and 8-line subway system and 10 thousand private vehicles. The experimental results show that our data-driven dispatching strategy significantly outperforms the state-of-the-art dispatching strategies without data-driven FHV insights.