Classification of Active Debt of Tax From the Sergipe State

• Yúri Faro Dantas de Sant’Anna Federal University of Pernambuco
• Luiz Felipe da Conceição Souza Federal University of Sergipe
• Antônio José Alves Neto Federal University of Sergipe
• Max Castor Rodrigues Junior Estácio University
• André Britto de Carvalho Federal University of Sergipe
• Renê Pereira de Gusmão Federal University of Sergipe

DOI: https://doi.org/10.5753/sbbd.2024.240204

Abstract

As a way of ensuring the efficiency of the services provided, the public sector depends on revenue from tax collection. In the State of Sergipe, the Goods and Services Circulation Tax (ICMS) is responsible for more than 90% of state revenue. Therefore, the effective collection of this resource is essential for state administration. Active debt is the name for debts that are judicialized and whose payment is still pending. Many taxpayers, in active debt, are unable to honor their debts or have debts considered insoluble. The costs generated by charges directed to these elements, in addition to the non-generation of revenue for the State through these processes, result in a series of additional expenses. This work proposed the evaluation and selection of a classification algorithm that, based on the taxpayer’s behavior and the characteristics of the debt, classifies the possibility of full, partial, or non-payment of the debt. After the proposed tests, the Random Forest algorithm presented the best performance among all the options explored, obtaining an f1-score of 81.84%. In this context, the classifier provides an efficient and viable solution for the real scenario.

Keywords: machine learning, classification, debt

References

Ahsan, M. M., Mahmud, M. P., Saha, P. K., Gupta, K. D., and Siddique, Z. (2021). Effect of data scaling methods on machine learning algorithms and model performance. Technologies, 9(3):52.

Arslan, M., Guzel, M., Demirci, M., and Ozdemir, S. (2019). Smote and gaussian noise based sensor data augmentation. In 2019 4th International Conference on Computer Science and Engineering (UBMK).

Brasil (1980). Lei nº 6.830, de 22 de setembro de 1980. dispõe sobre a cobrança judicial da dívida ativa da fazenda pública, e dá outras providências. Diário Oficial [da] República Federativa do Brasil.

Brasil (1996). Lei complementar nº 87, de 13 de setembro de 1996. dispõe sobre o imposto dos estados e do distrito federal sobre operações relativas à circulação de mercadorias e sobre prestações de serviços de transporte interestadual e intermunicipal e de comunicação, e dá outras providências. (lei kandir). Diário Oficial [da] República Federativa do Brasil.

Castellón González, P. and Velásquez, J. D. (2013). Characterization and detection of taxpayers with false invoices using data mining techniques. Expert Systems with Applications, 40(5):1427–1436.

Corvalão, E. D. et al. (2012). Classificação de contribuintes: um modelo em duas fases.

de Sant’Anna, Y. F. D., de Farias, M. L., Júnior, M. C., Dantas, D., and Junior, M. R. (2024). Fuel classification in electronic tax documents. In Proceedings of the 13th International Conference on Pattern Recognition Applications and Methods - Volume 1: ICPRAM, pages 337–343. INSTICC, SciTePress.

Duda, R. O., Hart, P. E., and Stork, D. G. (2001). Pattern Classification. Wiley, New York, 2 edition.

Gupta, M. and Nagadevara, V. (2007). Audit selection strategy for improving tax compliance: application of data mining techniques. In Foundations of Risk-Based Audits. Proceedings of the eleventh International Conference on e-Governance, Hyderabad, India, December, pages 28–30. Citeseer.

Guralnik, V. and Srivastava, J. (1999). Event detection from time series data. In Proceedings of the fifth ACM SIGKDD international conference on Knowledge discovery and data mining, pages 33–42.

Kubat, M. (2017). An Introduction to Machine Learning. Springer International Publishing, Gewerbestrasse 11, 6330 Cham, Switzerland, 2 edition.

Martín, H., Bernardos, A. M., Iglesias, J., and Casar, J. R. (2013). Activity logging using lightweight classification techniques in mobile devices. 17(4):675–695.

Oliveira, F. N. d. (2012). Estratégias para aperfeiçoar o processo de recuperação de receitas tributárias no estado da bahia: um modelo para o icms baseado em redes neurais artificiais.

Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., and Duchesnay, E. (2011). Scikit-learn: Machine learning in Python. Journal of Machine Learning Research, 12:2825–2830.

Portella, A. A. and Teixeira, R. P. (2016). Federalismo fiscal e efetividade da dignidade da pessoa humana: análise da posição do município na estrutura do financiamento público brasileiro e a escassez de recursos para as ações de saúde/fiscal federalism and effectiveness of human dignity... Revista de Direito da Cidade, 8(2):631–679.

Rocha, S. M. et al. (2017). Mineração de dados aplicada à classificação dos contribuintes de icms da sefaz-go.

SEFAZ-SE (2021). Contas anuais do estado 2021. [link]. Acesso em: 23 de junho de 2023.

Sisnando, S. R. A. and de Sousa Freitas, M. A. (2006). Previsão e avaliação do desempenho dos contribuintes do icms do estado do ceará utilizando as redes neurais artificiais. Revista econômica do Nordeste, 37(1):131–149.

Soares, L. E. and Oliveira, M. F. d. (2016). O protesto extrajudicial de certidão de dívida ativa como meio alternativo eficaz de recuperação do crédito público.