Metodologia para análise de mercado de ações com base em técnicas de aprendizado de máquina

Abstract

Predicting stock values is a difficult problem, financial time series have chaotic behavior and with the “mood” of the financial market being a determining variable for political and economic maneuvers many researchers have devoted their time to developing ways to predict their behavior. Using convolutional neural networks and long-term memory neural networks, we build models capable of predicting the price of the company Apple on the stock exchange. The results obtained with the proposed methodology indicate that neural networks are able to generalize the problem reasonably well, however the results obtained are not positive enough to justify the use of models in a competitive environment.