Enhancing Public Safety in Eswatini: A Machine Learning–Driven Predictive Policing Model

Résumé fourni par la source

Public safety remains a critical concern in Eswatini, as it prevents crime, reduces delayed response mechanisms, and optimizes police resources. This study applied machine learning techniques in predictive policing within the Kingdom of Eswatini (formerly Swaziland) to improve proactive law enforcement strategies and public safety. Crime has been a challenge in many societies and continues to threaten public safety, social cohesion, and economic development. Law enforcement agents often use reactive approaches to handle criminal incidents, which are generally associated with various impediments, such as delayed responses to crime incidents, resource‐intensive operations, victimization, and insufficient proactive crime prevention measures. Integrating machine learning techniques for predictive policing emerges as a new panacea for effective policing and crime prevention. However, there is a dearth of literature advocating proactive policing through predictive policing. Therefore, this study proposes a proactive approach to crime prediction and prevention by using machine learning models such as XGBoost, random forest, multilayer perceptron (MLP), and K‐nearest neighbors (KNN) models. These models were trained and tested using data from the Royal Eswatini Police Services (REPS). Our findings indicate that XGBoost provides the highest predictive accuracy at approximately 71.4%, with precision ranging from 0.65 to 0.81 and recall from 0.34 to 0.81, making it the preferred model for balanced performance across the metrics. Random forest recorded an accuracy of 66.2%, while MLP and KNN have 62.2% and 55.5% accuracy, respectively. The study recommends the integration of intelligence‐based models to enhance proactive crime prediction and identify potential crime hotspots. This can assist in optimizing resource allocation to prevent crime. Additionally, collaboration among stakeholders, including national security agents, policymakers, and the community, is essential to effectively adopt and utilize predictive policing technologies to enhance security operations.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.