APPLICATION OF RECURRENT NEURAL NETWORKS IN THE PROVISION OF TIME SERIES: A STUDY OF STOCK EXCHANGE STOCK PRICES

Abstract

Future events prediction in an assertive way has been the object of analysis by several researchers, especially in the financial area. In this context, there are countless possibilities for using this tool in the decision-making process of investment managers and analysts. This article aim to propose a recurrent neural network model based on the study of time series, oriented to the prediction and estimation of the price of shares on the Brazilian stock exchange. In this context, the present study enables the characterization of directions of financial trends and the prediction of prices from the training of the neural network, using real data from the Brazilian stock exchange in the year 2010. Regarding the methodology, the present research can be classified as applied, explanatory, in terms of its objectives, and quantitative in terms of its approach. The results obtained in this study reveal the ability to learn complex problems and, consequently, the possibility of application in other areas.

Keywords

• deep learning
• redes neurais recorrentes
• predição de preços
• bolsa de valores