Identifying Theft Hotspots: A Machine Learning Approach to Aggregated Data Analysis

Meyler, Mark (2025) Identifying Theft Hotspots: A Machine Learning Approach to Aggregated Data Analysis. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (729kB) | Preview

Abstract

Crime prediction is a critical domain of research due to its potential economic benefits and improvements in quality of life. Theft crimes represent a significant proportion of criminal incidents and, therefore, the ability to identify areas of increased theft risk, i.e. a hotspot, is essential to enable mitigation measures to be implemented by crime prevention agencies, but also to inform all interested parties of localised hotspots to safeguard against potential personal risk. A widely discussed challenge of effectively achieving this is the distinct lack of openly available crime data at low resolutions. This study proposes a geographically weighted LightGBM developed using fully openly available data at a coarse aggregation that can be used to predict theft hotspots at a more granular level. The paper focusses on burglary counts at MSOA levels to generate predictions at the finer LSOA scale, which are used to disaggregate the MSOA counts. Three different combinations of regions in England were utilised to evaluate scalability. The results demonstrate the enhanced capabilities of classifying areas of higher theft risks, with an increase in precision of up to 11% and recall of up to 20% in comparison to assuming the MSOA rate across all LSOAs within. There is some validation in the proposal that weighting the predictions from the local models can achieve more accurate and stable results, but further research is required to be conclusive. The integration of SHAP facilitates improved interpretability of key features contributing to these predictions.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Hasanuzzaman, Mohammed UNSPECIFIED
Uncontrolled Keywords:	Hotspot; Spatial disaggregation; coarse resolution; finer resolution; geographically weighted; spatial autocorrelation; interpretability
Subjects:	H Social Sciences > HV Social pathology. Social and public welfare > Criminology > Crimes and Offences Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions:	School of Computing > Master of Science in Data Analytics
Depositing User:	Ciara O'Brien
Date Deposited:	08 Sep 2026 10:21
Last Modified:	08 Sep 2026 10:21
URI:	https://norma.ncirl.ie/id/eprint/9883

Actions (login required)

View Item