Graphs are data structures that are practically used to learn complex systems of relations or interactions arising in a broad spectrum of problems ranging from biology and particle physics to social networks and recommendation systems.

In our research work, we attempt to gain the benefit of such a powerful data structure to better understand the contextual environment in which insured cars evolve. For this, we propose inferring the continuous hidden relationships that may exist between the different insured cars through an evolving graph that both captures the space and time variations. We exploit the constructed graph structure to devise a Spatio-Temporal Graph Network (STGN) to automatically learn novel features that simultaneously incorporate the spatial and temporal variation aspects.