AN INTEGRATED RISK PREDICTION AND MAXIMAL COVERING LOCATION PROBLEM MODEL FOR ALLOCATING SECURITY UNITS

This study proposes an integrated approach for the allocation of public security units, combining machine learning and optimization. A model based on XGBoost was used to estimate the risk of occurrence of Intentional Violent Deaths (IVD) across areas of the municipality of Caruaru, Brazil, using georeferenced historical data. The predictions were aggregated into a spatial risk index. Based on this index, an optimization model inspired by the Maximal Covering Location Problem was formulated to maximize covered risk under budget and personnel constraints. The model considers different types of units and allows for partial coverage and overlap. The results indicate that the approach is capable of identifying priority areas and guiding the efficient allocation of available resources, contributing to more informed and data-driven decision-making.