Self-exciting point processes and their applications to crime data

Self-exciting point processes describe a type of point process where the occurrence of an event leads to an increased probability of a future event taking place. Such subsequent events can be said to have been 'triggered' by the previous event. Originally used to model earthquakes, in recent years this class of processes has been utilised in crime modelling to describe how crime events can propagate future crime. In this thesis we look at ways of building on the existing literature which uses self-exciting point processes to model crime, with the intention of improving the ability to predict when and where future crime may occur.;We introduce a new parametric form for the triggering function, which can be utilised in situations where an initial event does not immediately lead to an increased risk of further events, but rather the risk increases over a period of time after the event. We also introduce adaptations to existing non-parametric methods which offer us fresh insight into the dynamics of crime. We validate our models using publicly available crime data from the city of Chicago.

Name dc:type.qualificationname

phd

Level dc:type.qualificationlevel

doctoral-pg

Grantor dc:publisher.institution

University of Strathclyde

Year dc:date.issued

2019

Author dc:creator

• Gilmour, Craig

Advisor dc:contributor.advisor

• Higham, Des

DOI

10.48730/0va1-bv17

Identifier

T15406

Author Identifier

201459674

OAI identifier oai:identifier

oai:strathclyde:f4752g788