Interactive cluster analysis of diverse types of spatiotemporal data

Abstract

We suggest an approach to exploratory analysis of diverse types of spatiotemporal data with the use of clustering and interactive visual displays. We can apply the same generic clustering algorithm to different types of data owing to the separation of the process of grouping objects from the process of computing distances between the objects. In particular, we apply the densitybased clustering algorithm OPTICS to events (i.e. objects having spatial and temporal positions), trajectories of moving entities, and spatial distributions of events or moving entities in different time intervals. Distances are computed in a specific way for each type of objects; moreover, it may be useful to have several different distance functions for the same type of objects. Thus, multiple distance functions available for trajectories support different analysis tasks. We demonstrate the use of our approach by example of two datasets from the VAST Challenge 2008: evacuation traces (trajectories of moving entities) and landings and interdictions of migrant boats (events).