Previsão de séries temporais no mercado financeiro de ações com o uso de rede neural artificial

• Ano de defesa: 2018
• Autor(a) principal: Carvalho, Valter Pereira de
• Orientador(a): Silva, Leandro Augusto da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Presbiteriana Mackenzie
• Palavras-chave em Português: rede neural; RNA; MLP; BOVESPA; Random Wall; previsão; mercado financeiro
• Área do conhecimento CNPq: CNPQ::CIENCIAS SOCIAIS APLICADAS::DIREITO::DIREITO PRIVADO::DIREITO CIVIL
• Link de acesso: http://dspace.mackenzie.br/handle/10899/24481

This work proposes a study of the forecast of time series with the use of data obtained from BOVESPA the basis of the values of the shares at the closing of the trading session. For the forecast, an arti_cial neural network (RNA) with MLP (MultiLayer Perceptron) architecture will be used. It will be shown through this prediction study of the financial market how the neural network behaves and how it can be of great value for forecasts with time series data. The analysis comprises the comparison between the forecast and the efective closing price within established periods. The paper compares the MLP network with the Random Walk Hypothesis. At the end of the study it is concluded that the artificial neural network used for stock market forecasting is able to show results very close to reality, and that this methodology can be used by individual and collective investors to understand the behavior of the actions and to orient themselves on the possible investment hypotheses.