This dissertation develops an intelligent, quantitative and probabilistic model to determine an optimal composition of a portfolio consisting of a financial asset and options over this asset. Initially we studied the characteristics of the historical distribution of returns and volatility of the most liquid stocks from the BOVESPA Stock Exchange, from January 2005 to July 2010, through a univariate and a bivariate polynomial regression. Characteristics such as mean reversion of volatility, strong correlation of historical and future volatility and a quadratic polynomial relationship between them were observed. A neural network was then developed to predict the future volatility of these stocks, since that is the most critical variable in determining an option´s price. Using the option pricing, we evaluated the performance of genetic algorithms in optimizing portfolios, structured with these derivatives, with three different evaluation functions in order to increase the potential return of investments while minimizing downside risks. The developed evolutionary system showed satisfactory results when the optimal portfolio was compared with several other market option strategies, demonstrating to be a relevant decision support system for investors and portfolio managers.