PREDICTIVE MODELING OF ROBBERIES IN CAMPO GRANDE (MS) USING SUPERVISED REGRESSION ALGORITHMS

This study investigates the application of supervised regression algorithms for the monthly prediction of robberies in the city of Campo Grande (MS), based on operational records from the Military Police of Mato Grosso do Sul (PMMS) covering the period from 2018 to 2023. The adopted methodology consisted of comparing various machine learning techniques, including Linear Regression, Ridge, Lasso, Decision Tree, Random Forest, and Multilayer Perceptron (MLP), with traditional statistical methods relying on means and medians. The objective was to assess whether supervised models outperform simplified baseline approaches. The dataset, composed of monthly aggregated records for the entire municipality, reliably reflected the behavior of preventive policing actions and their relationship with reported robberies. Model evaluation was conducted using MAE (Mean Absolute Error), MSE (Mean Squared Error), and R² (Coefficient of Determination). Empirical validation with real data from 2023 demonstrated the superiority of the supervised learning approach, with Random Forest achieving the lowest MAE (10.43) and the MLP model achieving the lowest MSE (174.04) and highest R² (0.44). The findings highlight the potential of supervised algorithms as decision-support tools in public security, promoting a data-driven approach to management and enhancing the ability to anticipate critical scenarios. Keywords:Crime Prediction; Supervised Regression; Public Security; Machine Learning; Predictive Policing.