| Abstract |
In developing effective public security policies, two aspects are essential: intelligence and information. In this context, the classification of Macrocauses can help governments better understand some aspects of public security related to the analysis of violent crimes in a state. In the Brazilian state of Rio Grande do Norte (RN), these analyses are currently done by analyzing raw data done by an expert without any specific auxiliary system, known as traditional approaches. It is a time-consuming task, and different experts may have different views on the same case. With this in mind, in this work, we intend to assist these criminal experts by providing a predictive model that classifies the Macrocauses of crimes into a set of pre-defined types to guide the work of the experts and, as a consequence, to speed up the analysis and management processes. Thus, a feature preprocessing was performed in which three different classification techniques are analyzed. The results show that when using XGBoost, the model accuracy reached 0.932378, which is considered excellent for this application. © 2021 IEEE. |
