A semi-parametric spatiotemporal Hawkes-type point process model with periodic background for crime data

Criminologists try to predict crime with a number of methods, such as “hot-spotting” – making maps of locations where crimes tend to occur – and epidemiological techniques based on the assumption that the local risk  of crime rises temporarily after a crime occurs. More recently, researchers have turned to modelling with Hawkes-type point-processes, which can incorporate time-varying hot spots. However, such studies have neglected periodic influences expected to be important for crime. The behaviour of criminals, as human beings, should reflect their biological rhythms and as well as the daily and weekly periodic patterns of social activity.
In a recent paper, LML Fellow Jiancang Zhuang and Jorge Mateu analyse crime data using an extended  Hawkes model. Using data for violent crimes and robberies in Castellon, Spain, over a two-year period, they focus on disentangling the periodic components from the long-term trend in the background rate. Their results indicate that the robbery crime is highly influenced by the daily and weekly rhythms of social life, and that about 3% of such crimes can be explained by clustering. They also show that the model can likely be improved by distinguishing daily occurrence patterns on weekends and working days, and noting that robbery activity near the city centre shows different temporal patterns from suburban areas.
The paper is available at http://bemlar.ism.ac.jp/zhuang/pubs/zhuang2018jrssa-manuscript.pdf
LML is a charity. If you enjoyed reading this, please consider supporting us.

Leave a Reply

Your email address will not be published. Required fields are marked *