Spatiotemporal forecasts of London’s crime hotspots

McMorrow, Andrew (2022) Spatiotemporal forecasts of London’s crime hotspots. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (2MB) | Preview
Preview	PDF (Configuration manual) Download (211kB) | Preview

Abstract

This study investigates and predicts future crime hotspots of London via spatiotemporal modelling. A model was sought which provided ethical, unbiased and accurate predictions of forecasted crime hotspots in London, with the widespread societal benefits of reduced crime rates serving as motivation. This was successfully achieved via construction of the Extreme Gradient Boosting (XGB) model to predict future crime rates, before crime hotspot forecasts over the period of one year were made on these via the use of quantile statistics. This model design was deemed the most appropriate in terms of the aforementioned requirements due to its added interpretability and favourable results achieved in comparison to literature. The data used consisted of crime data in the Greater London Area (GLA) from 2012-2021, along with relevant spatial such as the number of food outlets, nightlife spots, entertainment venues etc. in a given area. Crime data in the GLA from 2012-2021 was used for this modelling, along with relevant spatial such as the number of food outlets, nightlife spots etc. in a given area. The GLA was partitioned into equi sized grid cells, with the data per cell aggregated over a given year into each cell before implementation into the model. The Prediction Accuracy Index (PAI) (formulated by Chainey et al. (2008)) was deemed the most appropriate performance metric to base results upon. Predictions for hotspots were made on unseen data for both before and during the COVID-19 pandemic, with the most appropriate model returning PAI scores of 2.84 and 2.97. This offered an improvement of 11%/22% on pre/mid-pandemic data over the highest performing model which did not incorporate the data on spatial features. This consistent performance despite wholesale changes in crime rates further indicates that this model is fit for use to accurately and ethically predict future crime hotspots in London.

Item Type:	Thesis (Masters)
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science H Social Sciences > HV Social pathology. Social and public welfare > Criminology H Social Sciences > HT Communities. Classes. Races > Urban Sociology
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Tamara Malone
Date Deposited:	23 Feb 2023 10:25
Last Modified:	23 Feb 2023 10:25
URI:	https://norma.ncirl.ie/id/eprint/6224

Actions (login required)

View Item