A Graph-Based Approach to Spatiotemporal Event Sequence Mining
Berkay Aydin, Rafal A. Angryk · 2016
Sequential pattern mining from spatiotemporal data has received much attention in recent years due to its broad application domains such as targeted advertising, location prediction for taxi services, and urban planning. The characteristics of spatiotemporal sequences vary widely depending on the discovered knowledge type. Most of the recent approaches focus on the point-based spatiotemporal data presumably because of its greater availability. However, the region-based spatiotemporal data, primarily obtained from scientific resources, has not received much attention. In this work, we introduce an algorithm for mining spatiotemporal event sequences (STESs) from trajectory-based event instances. We consider each instance to be associated with an event type. We propose a graph-based mining algorithm, which transforms the sequences of spatiotemporal trajectories into a directed acyclic graph, and discovers the frequently occurring sequences of event types. Our proposed algorithm adopts a pattern-growth based approach utilizing the directed edges from the graph and discovers the event sequences without expensive candidate generation steps.