Precificação de opções de dólar no mercado brasileiro utilizando redes neurais e algoritmos genéticos

• Ano de defesa: 2007
• Autor(a) principal: Chagas, Guido Marcelo Borma
• Orientador(a): Rochman, Ricardo Ratner
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Palavras-chave em Português: Derivatives; Genetic algorithms; Derivativos; Algoritmos genéticos; Redes neurais
• Palavras-chave em Inglês: Options; Neural networks
• Link de acesso: http://hdl.handle.net/10438/2055

This work compared, under usual macroeconomic conditions, the effectiveness of the Neural Networks (NN) model enhanced by Genetic Algorithms (GA) in Dollar options’ valuation with the following conventional valuation models: Black-Scholes, Garman-Kohlhagen, Trinomial Trees and Monte Carlo Simulations. All information employed in this analysis, comprehended between July, 1999 and December, 2006, was provided by Bolsa de Mercadorias e Futuros (BM&F) and by Federal Reserve. Comparisons and assessments were conducted with the MATLAB software, version 7.0, and its toolboxes which provided the necessary tools and environment to develop and implement the models previously mentioned. The delta-hedging cost’s analyses of each model indicated that, even though more complex, the use of Genetic Algorithms to directly optimize (i.e., at binary level) the Neural Network’s synaptic weights did not produce any significantly superior results than the conventional models.