Prototyping A Web-based High-Performance Visual Analytics Platform for Origin-Destination Data

Abstract

Origin-Destination (OD) data has quickly emerged as a popular and fast growing spatiotemporal data type due to widely adoption of GPS, smartphones and location dependent social networks. Several previous works have developed techniques for managing and visualizing large-scale OD data in desktop computing environments. In this study, by leveraging our experiences in Web-GIS and parallel spatial data processing and learning from successful OD data visualization case studies, we have developed a Web-based high-performance visual analytics prototype platform for OD data. Observing that interactive spatial queries typically only involve a limited number of Regions of Interests (ROIs), we propose a simple yet effective technique to aggregate OD records into dynamically defined OD polygons by data parallel scanning OD point locations for cache efficiency and easy parallelization on conventional multi-core hardware for high efficiency and performance. By dynamically integrating with a graph database backend, our prototype platform is capable of visualizing social network analytical results and guide users to further retrieve detailed information of interests. Two experiments are provided to demonstrate the utilization of the proposed techniques, including web frontend functionality and backend efficiency, by using more than 170 million taxi trip records in NYC in 2013 as well several urban infrastructure datasets. Interactive demonstrations are available for the web-based system