New Zealand Statistical Association 2024 Conference

Charlotte Jones-Todd

Charlotte: University of Auckland; Conor: University of Otago

Accounting for social networks in a self-exciting point process model

This is joint work with Conor Kresin

A Hawkes process models self-exciting behaviour in temporal point pattern data. Self-excitement is where the occurrence of an event increases (i.e., excites) the chance of another in quick succession. This phenomenon is apparent in many crime-based event data (e.g., gun and knife violence). In addition to the occurrence of these crime events there are often known links between them. For example, crimes may be linked due to 1) the same perpetrators, or 2) co-perpetrators of another crime being involved. This network of linked crimes evolves and expands over time (e.g., via co-perpetration) and distinct clusters of events form. We develop a Hawkes model that incorporates time-evolving social network dynamics accounting for the dependency between events linked by a given social network.