Passenger Mobility Analysis based on Car-hailing Platform Data and Artificial Intelligence Algorithms

The ride-hailing service platforms have grown tremendously around the world and attracted a wide range of research interests. This talk will introduce some recent progress we have made on two key issues of ride-hailing service: demand forecasting and order matching. A key to ride-hailing service platforms is how to realize accurate and reliable demand prediction. However, most of the existing studies focus on region-level demand prediction while only a few attempts have been made to address the problem of origin-destination (OD) demand prediction. In our recent studies, we constructed dynamic OD graphs to describe the ride-hailing demand data from a graph perspective. We proposed two novel neural architectures named the Dynamic Node-Edge Attention Network (DNEAT) and the Dynamic Auto-structuring Graph Neural Network (DAGNN) to address the unique challenges of OD demand prediction from the demand generation and attraction perspectives. Online matching between idle drivers and waiting passengers is another key component in a ride-sourcing system. We proposed two ideas to improve the matching efficiency: early driver matching and delayed passenger matching. The first idea considers the vehicles whose destinations are close to the passenger’s origin, so that the passenger’s waiting time may be shorter and the vehicle’s pick-up distance and fuel consumption can be saved. The second one assumes that a specific passenger request can benefit from a delayed matching since he/she may be matched with closer idle drivers after waiting for a few seconds. The problems are solved by reinforcement learning methods. Through extensive empirical experiments with well-designed simulators, we show that the proposed frameworks can remarkably improve system performance.