Modelo de média-variância de Markowitz em conjunto com modelo ARIMA e inteligência artificial para maximização de resultados financeiros em operações aplicável em engenharia

Abstract

Expectations about the future outcome of markets and companies are largely responsible for their quotations. The form of variation of the data involved can be considered complex, since several methods are developed in order to try to make predictions and, at the same time, generate greater returns in investments such as variable income, a subject that in recent years has awakened the interest of an increasing portion of the population. In the financial market, the possibility of obtaining rates of return above the rates obtained in fixed income is one of the main factors that lead to the migration of resources from fixed income to variable income. This greater return is linked to a greater risk assumed by the investor, and using mathematical knowledge to guide decision-making is highly recommended. This work aimed at the study and application of Artificial Neural Networks (ANN) and the integrated autoregressive moving averages model (ARIMA), as well as the Markowitz Mean-Variance model in the Brazilian capital market, with the objective of maximizing the creation of value to the investor by controlling the risk of the portfolio and predicting the price of its assets. For this, asset selection criteria were used, to then predict the quotation of listed companies and later apply the Mean-Variance model to calculate the participation percentages of each asset, thus creating the Optimized Portfolio. For this prognosis, a comparison of time series forecasting models was carried out in 243 trading days of the Brazilian stock exchange (B3) in the year 2021, using RNA and ARIMA. Previous results show that the methodology is capable of generating financial returns and can be applied to similar cases. [resumo fornecido pelo autor]