Abstract:

With the rapid development of global positioning systems (GPS), smart mobile devices, and sensing technologies, a huge amount of data with geo-location and time information has been collected. Such data, commonly referred to as Spatio-Temporal Big Data (STBD), has the potential to transform the society. However, STBD poses significant research challenges for data analytics and mining such as data-variety and candidate pattern cardinality exceeds the capacity of current spatial computing analytics and management techniques. This talk presents examples of new STBD analytics and mining techniques to improve urban sustainability. Specifically, the main part of the talk will focus on mining GPS trajectories of vehicles to early detect urban gathering events. This problem is of great importance to reducing congestion and public safety risks (e.g., stampedes). Prior techniques on event detection only find events with undirected footprints (e.g., circles, rectangles, or undirected networks), therefore lacking the ability to characterize how traffic gather towards the destination. By contrast, our work proposes a novel concept called Gathering Graph (G-Graph) to model the destination as well as the gathering paths of a gathering event. We propose a smartEdge algorithm to efficiently discover the most “interesting” G-Graphs for each time slot. Case studies and simulations on real taxi GPS data show that the proposed technique can identify gathering events accurately and timely. Experiments on runtime efficiency are also provided. The talk will also discuss other problems such as analyzing taxi trip data to recommend taxi driving directions for better business efficiency.