Data-Driven Modeling of Ride-Hailing Trajectories

Abstract

Predicting the drivers' movements in a spatial ride-hailing network accurately and promptly is essential for developing both ride dispatching and dynamic pricing algorithms. In this paper, we construct a data-driven model based on recurrent neural networks. We test the performance of our model using a real-world trajectory database. We find that our approach outperforms the base model (with an accuracy of 74%~78%). We also conclude from the analysis that drivers only look ahead a short distance when making route choice decisions. This model is useful for developing operational algorithms, traffic simulators, and qualitative studies.