Title:T-PickSeer: Visual Analysis of Taxi Pick-up Point Selection Behavior

Abstract:Taxi drivers often take much time to navigate the streets to look for passengers, which leads to high vacancy rates and wasted resources. Empty taxi cruising remains a big concern for taxi companies. Analyzing the pick-up point selection behavior can solve this problem effectively, providing suggestions for taxi management and dispatch. Many studies have been devoted to analyzing and recommending hot-spot regions of pick-up points, which can make it easier for drivers to pick up passengers. However, the selection of pick-up points is complex and affected by multiple factors, such as convenience and traffic management. Most existing approaches cannot produce satisfactory results in real-world applications because of the changing travel demands and the lack of interpretability. In this paper, we introduce a visual analytics system, T-PickSeer, for taxi company analysts to better explore and understand the pick-up point selection behavior of passengers. We explore massive taxi GPS data and employ an overview-to-detail approach to enable effective analysis of pick-up point selection. Our system provides coordinated views to compare different regularities and characteristics in different regions. Also, our system assists in identifying potential pick-up points and checking the performance of each pick-up point. Three case studies based on a real-world dataset and interviews with experts have demonstrated the effectiveness of our system.