Classificação da percepção de servidores públicos federais em relação a atos de corrupção utilizando algoritmos de aprendizado de máquina

• Autor(a) principal: Ariza, Vinicius Matheus Pimentel
• Data de Publicação: 2023
• Outros Autores: Santos, Bruno Samways dos
• Tipo de documento: Artigo
• Idioma: por; eng; spa
• Título da fonte: Brazilian Journal of Production Engineering
• Texto Completo: https://periodicos.ufes.br/bjpe/article/view/42073
• topic: Mineração de Dados; Corrupção; Serviço Público Federal; Data Mining; Corruption; 5W2H; Federal Public Service; Minería de Datos; Corrupción; Servicio Público Federal

Computational techniques have proven useful in the fight against corruption in the public sector, enabling the early detection of suspicious activities. The aim of this study was to compare machine learning algorithms in the context of observing acts of corruption in the Public Service. In this regard, data extracted from a survey conducted by the World Bank in 2021 on the topic of Ethics and Corruption in the Public Service were analyzed, involving approximately 22.000 respondents. The development of models aimed at promoting transparency and integrity in the Brazilian public service is proposed. The results demonstrated the feasibility of using machine learning techniques, with Logistic Regression proving to be the best option for the studied scenario, with an accuracy of 82%. The developed model and generated analysis can be used to assist in the identification of suspicious corruption activities in the public sector, contributing to early detection and prevention of illegal practices. The results also highlight the importance of developing public policies to promote ethics and integrity in public service, as well as the role of advanced technologies in improving governance and society's trust in public institutions.