Year: 2025
Type: article
Abstract: Data analysis is crucial for optimizing strategies in various sectors, with Public Safety being an area of great potential. This work analyzes crime data from the city of São Luís, Maranhão, from 2020 to 2023, provided by the Integrated Police Operations Center (IPOC), with the goal of developing a high-performance classifier to predict serious crimes. The methodology involved rigorous preprocessing of a database with approximately 52,000 records, applying techniques to address missing values and balance crime classes using the SMOTE (Synthetic Minority Over-sampling Technique) method. Six Machine Learning algorithms were trained and evaluated: Decision Tree, K-Nearest Neighbors (KNN), Multi-Layer Perceptron (MLP), Support Vector Machine (SVM), Random Forest, and Naive Bayes. The results demonstrated the superiority of the Random Forest and SVM models, which achieved perfect accuracy, precision, and F1-score metrics (1.0000), indicating high generalization capacity and ability to identify criminal patterns. The study concludes that the application of these models represents a robust tool to support police planning and public policy formulation, aiming to reduce crime in high-risk areas, such as the city's East Quadrant, which concentrates the highest number of incidents.
Source: Caderno Pedagógico
Authors: Felipe de Oliveira Carvalho, Silvio Carlos Leite Mesquita, Marta de Oliveira Barreiros, José Artur Lima Cabral Marques, Jonathan Araujo Queiroz, Allan Kardec Duailibe Barros Filho
Institution: Universidade Federal do Maranhão
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W4414495357