Previsão de curto prazo de receitas tributárias em séries temporais com uso de redes neurais

Abstract

The Bahia State Finance Department's mission is to provide and manage the financial resources that make public policies of the State viable Bahia, being responsible for tax collection and financial management of state accounts. As a result, and given the impact that this control has on state finances, monitoring the behavior of the flow of income and expenses (cash flow) is of vital importance to this secretariat. Constituting 80% of the revenues of the State of Bahia (BAHIA, 2020), the tax revenues, with emphasis on ICMS1, are critical for the execution state financial institution, which depends on them for the provision of its services. of this way, it is vital that an estimate can be obtained, with a good precision and accuracy, of the total value of tax revenues that will be collected by the taxpayers in the next few days. With this forecast, the Financial Area of The Registrar's Office can make important decisions regarding the financial transactions of the Caixa Único do Estado, being able to decide better on payments and/or financial investments of the resources available on the days following. Currently, this collection forecast is made using methods traditional statistics, with the use of spreadsheets and queries to historical data. This way of making this prediction lacks greater efficiency and precision, demanding, still, dedication of specialized personnel for its elaboration. This leaves a gap to be filled, in the form of a solution that is able to automatically and reliably provide the expected value for the total amount of State tax collection in the five future days. Such a solution has the potential to provide a great benefit to the management of finances by the Secretary of Finance, allowing a better control of its cash flow with lower labor cost involved in this forecast