|Principal Investigator||Yu Zhang, Ph.D.|
|Final Report (DOI)||Available Soon|
|Policy Brief||Available Soon|
|RIP||View RIP entry|
Taxi and Transportation Network Company (TNC) are important to the urban transportation system. An accurate short-term forecasting of passenger demand can help operators better allocate taxi or TNC services to achieve supply-demand balance in real-time, so drivers can improve the efficiency of picking up passengers and passengers can reduce waiting time. Previous research of demand forecast has proposed sophisticated machine learning and neural network-based models to predict short-term demand for taxi or TNC service; however, few of them jointly model these two modes although the short-term demand of taxi and TNCs is closely related. With Uber started to include NYC taxi on their platform, capturing information-sharing between the two modes could potentially reduce prediction error for both. Thus, this study proposes a multi-task learning (MTL) model that jointly predict the shortterm demand of both modes. In addition, graph neural network will be applied to capture the spatial correlation of demand. The study will perform experiments with different MTL models and techniques for modeling spatial correlation of demand for the case study of Manhattan, New York City. The prediction accuracy of single-task learning and multi-task learning models will be compared and future research direction will be discussed.