Spatio-temporal pattern detection in spatio-temporal graphs

Abstract

Spatio-temporal (ST) graphs have been used in many application domains to model evolving ST phenomenon. Such models represent the underlying structure of the phenomenon in terms of its entities and different types of spatial interactions between them. The reason behind using graph-based models to represent ST phenomenon is due to the existing well-established graph analysis tools and algorithms which can be directly applied to analyze the phenomenon under consideration. In this paper, considering the use case of two distinct, highly dynamic phenomena - invasive team sports, with a focus on handball and urban road traffic, we propose a spatio-temporal graph model applicable to both these phenomena. Different types of entities and spatial relations which make up these phenomena are highlighted to formalize the graph. Furthermore, the idea of graph-based pattern detection in both these phenomena is explored. Different types of ST patterns for both ST phenomena are discussed and the problem of pattern detection is formalized as the problem of subgraph isomorphism for dynamic graphs. Finally, the results of our algorithm to detect random ST patterns in random ST graphs are presented. The ideas discussed in this paper are applicable to other ST phenomena as well.