Classificação de empresas em grupos de retorno das ações utilizando redes neurais artificiais

Abstract

The application of artificial neural networks focused on the financial market has been used in several studies in the area of Finance and presented good results for the purpose of forecasting and classification. On the other hand, financial managers and investors use the numbers found in the financial statements to reflect the future situation of the company and make decisions. Based on these evidences, the objective of this study was to classify the return of shares of Brazilian publicly traded companies based on quarterly accounting, financial and market indicators from 2015 to 2017 processed in Neural Networks. The specific objectives were to identify the variables that influence the return of the organizations' actions. A factorial analysis was then applied to indicators to better group them into factors and through SPSS software an optimized RNA was run, having the output of the quarterly return of shares divided into quartiles and the input of all other factors. The results showed that the RNA used in the study was able to properly classify 100% of the companies with returns of the first and fourth quartiles, showing the effectiveness of the technique to segregate companies with higher and lower shareholder returns from indicators.