Problemas inversos em engenharia financeira: regularização com critério de entropia

• Autor(a) principal: Ramilijaona, Raombanarivo Dina
• Data de Publicação: 2013
• Tipo de documento: Dissertação
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da UERJ
• Texto Completo: http://www.bdtd.uerj.br/handle/1/13788
• topic: Inverse problems; European call option; Binary option; Entropy; Exponential; Newton-Raphson; Implicit volatility; Problemas inversos; Opção europeia; Opção binária; Entropia; Exponencial; Newton-Raphson; Volatilidade implícita; Investimentos Modelos matemáticos; CNPQ::CIENCIAS EXATAS E DA TERRA::MATEMATICA::MATEMATICA APLICADA

This study aims at applying Maximum Entropy Regularization to the Inverse Problem of Option Pricing suggested by Neri and Schneider in 2012. They pointed out that the probability density that solves such problem in the case of calls and digital options could be written as piecewise exponentials on the positive real axis. The limits of these segments are the different strike prices. The entropy criteria is a powerful tool to regularize this ill-posed problem. The Exponential Family solution set is calculated using a Newton-Raphson algorithm, with specific bounds for the binary options. These bounds obey the no-arbitrage principle. We applied the method to data from the Brazilian stock index BOVESPA and its call prices for different strikes. The parametric entropy analysis for "synthetic" digital prices (constructed from the no-arbitrage bounds) showed values where the digital prices maximizes the entropy. The example of data extracted on the IBOVESPA of January 24th 2013, showed slippage from the no-arbitrage principle when the option was in the money: such principle is a necessary condition to apply the maximum entropy regularization to get the density and modeled prices. When the condition is fulfilled, our results showed that it is possible to have a smile-like volatility curve with prices calculated from the exponential density that fit well the market data. In a computational modelling perspective, this thesis enabled the implementation of a pricing method using the maximum entropy principle. Three well known algorithms were used in that extent. The bisection alone, then a combined bisection with Newton-Raphson to recover the implied volatility from market data. Thereafter, the one dimensional Newton-Raphson to calculate the coefficients of the exponential densities: purpose of the study. Finally the Simpson method was used to calculate integrals of the cumulative distributions and the modeled prices implied by the expectation.