Aplicação de machine learning na classificação de restos a pagar

Resumo

The Federal Public Budget is a formal instrument (Annual Budget Law – LOA), with an annual periodicity, which contains estimates of revenues and authorization of expenses to be incurred in the respective year. Despite the budgetary principle of annuality, the Remainders to be Paid provide relative flexibility to the budget, allowing the Manager to conclude the execution of public expenses in years subsequent to the LOA. This concession of additional time allows for a greater and more suitable execution time for the expenses arising from the contracting of works, however, the cancellation of the expenses entered in the condition of remains to be paid translates into budgetary losses due to the restrictions on the reallocation of the credit authorized in the LOA . Machine learning is a field of Artificial Intelligence that encompasses the study and construction of algorithms capable of autonomously acquiring knowledge. These computational techniques analyze a large amount of data in order to automatically identify patterns based on examples, building a learning model. Our purchases with credit cards are inputs for the machines to learn what we want to consume. Therefore, the commitments of works in the process of registering outstanding amounts can be inputs for these algorithms to learn which ones are likely to be canceled. The objective of the study was to carry out a simulation of the registration process of remains to be paid, classifying the works contracted by the Directorate of Military Works of the Brazilian Army, in the period from 2016 to 2018, using Machine Learning. The results obtained with the application of the Random Forest algorithm indicated a significant reduction in percentage terms and absolute values of cancellations of remaining payable of the analyzed expenses. In the Federal Public Administration, approximately 10% of the values registered in outstanding amounts payable, referring to the contracting of works, are canceled by the managers, constituting a relevant amount, in the order of half a billion reais, justifying the importance of research in the contribution of the improvement of public management in the allocation of resources authorized in the Annual Budget Law.

Ficha do documento

Tipo

Dissertação

Ano

2023

Instituição

Fundação Getulio Vargas

Fonte

Repositório da FGV

Idioma

Português

Acesso

Acesso aberto

Identificador

oai:repositorio.fgv.br:10438/33359

Temas

EconomiaGestão PúblicaFinanças PúblicasAdministração PúblicaTecnologia

Palavras-chave

Orçamento públicoDespesas públicasInteligência artificialRestos a pagarMachine learningRandom forestPublic budgetPublic expenditureRemains payableEconomiaOrçamento - BrasilDespesa públicaAprendizado do computador