Application of GIS in crime prediction: a case study of burglary within Nakawa Division, Kampala District

Abstract

Crime is a common social problem that affects the quality of life and the economic growth of a society. Whether causing physical harm, psychological trauma or economic setbacks, crime affects everyone. In most developing countries, crime management strategies are largely based on the rate at which crimes are reported at a police station. Little effort has been put in crime intelligence and prediction to detect and prevent crimes before they occur. As a result, crime responses put a huge financial burden on taxpayers and governments. This study therefore aimed at exploring the application of GIS and crime theories as an alternative measure to effectively predict crime for better crime management. In this study, burglary and the Social Disorganization theory were selected as the crime and theory respectively. The study was limited to Nakawa Division in Kampala District. Crime data for the year 2019 was obtained from Jinja road police station. Datasets identified from literature review included; number of facilities, percentage of working status, percentage of indecent dwelling, police stations and population density. The data was tested for spatial patterns and bi-variate dependencies using Moran’s I and scatter plots. The burglary data was found to be clustered while no significant spatial dependencies were observed between the datasets, meaning the datasets were appropriate for further analysis. The other dataset also showed no correlation among themselves. The data sets were then run in an iterative Ordinary Least squares Regression procedure to identify the key variables that can explain burglary incidences in the division. Basing on the derived coefficients of the variables, police stations and population density were eliminated from the final regression using Geographically Weighted Regression. GWR results generated an approximately 60% R2 value, meaning that the model can predict burglary incidences at 60% confidence level. The results therefore concluded that integration of GIS and crime theories can provide an alternative measure for crime prediction. The performance of model was however affected by generalized census data and the spatial unit(parishes). Further to this, Bukoto I, Bukoto II and Kyanja had little or no data. The study finally recommended that more independent variables, a smaller spatial unit and a wider time span for the burglary data be adopted for better burglary predictions.