Exploring Global Gun Violence Prediction through Machine Learning Models

Abstract

Gun violence is a pressing global concern, inflicting dire consequences on individuals, communities, safety, and healthcare systems. This study uses correlation analysis to investigate the relationships between key factors, such as gun regulations, firearms per capita, and registered and unregistered firearms. It uses correlation analysis and feature coefficients, offering a more detailed understanding of how each variable influences the prediction of the prevalence of gun violence. The study employs machine learning models, revealing noteworthy disparities in their performance. Notably, the Logistic Regression model achieved a modest accuracy score of 64.7 per cent, mainly due to its linearity. In contrast, the Decision Tree and Random Forest models outshone the Logistic Regression model, achieving impressive accuracy scores of 95.6 and 96.1 per cent, respectively. These results highlight intricate relationships, including the correlation between stricter gun regulations and lower predicted gun-related death rates. Moreover, positive correlations between incidents such as 'deaths by firearm' and higher predicted rates emphasise the predictive potential of Decision Trees and Random Forests, offering a promising approach to mitigating gun violence and identifying its root causes. This study advances our comprehension of the factors influencing gun violence and the potential impact of legislative measures in addressing this urgent global concern.