Use of Econometric Predictors and Artificial Neural Networks for the Construction of Stock Market Investment Bots

• Published in: Computational economics Vol. 61; no. 2; pp. 743 - 773
• Main Authors: Nametala, Ciniro A. L., Souza, Jonas Villela de, Pimenta, Alexandre, Carrano, Eduardo Gontijo
• Format: Journal Article
• Language: English
• Published: New York Springer US 01-02-2023; Springer; Springer Nature B.V
• Subjects: Artificial intelligence; Artificial neural networks; Behavioral/Experimental Economics; Brokerage; Computer Appl. in Social and Behavioral Sciences; Econometrics; Economic Theory/Quantitative Economics/Mathematical Methods; Economics; Economics and Finance; Financial services industry; Forecasts and trends; Math Applications in Computer Science; Neural networks; Operations Research/Decision Theory; Operators; Political crises; Research subjects; Securities markets; Stock exchanges; Stock markets; Stocks; Volatility; Time series forecasting; Quantitative finance; Econometrics; Optimization; Artificial neural networks
• ISSN: 0927-7099; 1572-9974
• DOI: 10.1007/s10614-021-10228-0

The gradual replacement of human operators by investor bots in financial markets has changed the way assets are traded on stock markets. The use of strategies based on artificial intelligence has became a hot trend, especially due to improvements in computer processing capacity. Recent works showed that the adoption of strategies based on classical predictors has decreased considerably over the past years. In this sense, hybrid approaches, which combine artificial intelligence and classical predictors, have risen as an interesting research subject. This manuscript introduces an investor bot that combines predictions made by two classes of artificial neural networks and three classes of econometric predictors. Such a combination are accomplished by ensembles, which are continuously re-optimized with the intention of identifying profitable opportunities. Data from the Brazilian stock market, with daily granularity, was used. Models generated from such data were applied in a period of economic and political crisis, which leaded to the fall of the Brazilian international investment rating. Superior results were obtained against the benchmarks, despite the brokerage costs and high-price volatility in such a period.