Análise dos fatores que contribuem para a recuperação dos créditos de ICMS inscritos em divida ativa no estado do Rio de Janeiro

• Autor(a) principal: Guimarães, Tiago Rodrigues
• Data de Publicação: 2023
• Outros Autores: [e-mail]
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/20528
• topic: XGBoost; Active Debt; Public Management; Machine Learning Algorithm; Credit Rating.; Dívida Ativa; Gestão Pública; Algoritmo de aprendizado de máquina; Classificação de créditos; CIENCIAS SOCIAIS APLICADAS::ADMINISTRACAO::ADMINISTRACAO PUBLICA::CONTABILIDADE E FINANCAS PUBLICAS

Prosecutor's Offices are generally known for their bureaucratic approach, following formal procedures as legally prescribed, but often without evaluating the efficiency and effectiveness of their actions. This leads to a backlog of tax executions in the judicial system, causing a negative impact on public funds as efforts are not always proportionally rewarded. In light of this context, the objective of this study is to contribute to the improvement of the management of the State of Rio de Janeiro's active debt. To achieve this, an approach is proposed that provides a classification of the probability of recovering the registered credits, allowing the State Prosecutor's Office of Rio de Janeiro (PGE/RJ) to allocate recovery efforts more efficiently while recognizing these credits more objectively in accounting terms. The methodology adopted in this study involves quantitative analysis and the use of the machine learning algorithm known as XGBoost. Through this approach, factors influencing the probability of recovering Active Debt Certificates (CDA) will be identified, considering variables such as days in arrears, total debt amount, days of company existence, registration status, and region, among others. Based on the obtained results, a credit rating will be established, providing a clearer view of the probability of recovery for each CDA and allowing for the prioritization of recovery efforts in cases with a higher likelihood of success. Additionally, these results can contribute to a better estimation of losses related to the State of Rio de Janeiro's active debt. Currently, PGE/RJ relies on a method based on professional judgment of prosecutors, without the use of objective statistical criteria, to estimate losses from active debt credits. However, based on the results of the proposed model, it is estimated that the losses amount to approximately 98,64% of the total active debt value, highlighting the need for a revision of this methodology. Therefore, the results obtained through the XGBoost can significantly contribute to a more accurate estimation of losses from the State of Rio de Janeiro's active debt credits. It is noteworthy that the developed model demonstrated high performance in predicting credit recoverability. Moreover, this more objective and data-driven approach can assist PGE/RJ in directing recovery efforts more effectively, resulting in improved management of the active debt and financial returns to public funds.