APPLICATION OF MACHINE LEARNING TECHNIQUES TO PREDICT HOTSPOTS OF INTENTIONAL VIOLENT DEATHS IN THE MUNICIPALITY OF CARUARU, INTERIOR OF PERNAMBUCO.

This article presents a predictive model for identifying hotspots of Intentional Violent Deaths (MVI) in the municipality of Caruaru, in the interior of Pernambuco, through the application of machine learning algorithms to temporal and geospatial data. The data used refer to real cases of MVI in the municipality, and cover the period from 2017 to 2024. The modeling was conducted with different machine learning algorithms, using techniques such as hyperparameter optimization and class balancing. The results demonstrated satisfactory performance, with emphasis on the XGBoost model, with an average F1-score of 75% and recall of 80% in the monthly prediction of hotspots in 2024. Approximately 26% of the homicides registered in the year occurred in areas previously identified as hotspots, highlighting the potential of the approach as a preventive tool.