American option pricing using Bayesian multi-layer perceptrons and Bayesian support vector machines

摘要

An option is the right, not the obligation, to buy or sell an underlying asset at a later date but by fixing the price of the asset now. There are European and American styled options. European styled options can be priced using the Black-Scholes equations but American options are more complex and valuable due to the second random process they introduce. Multi-layer perceptrons and support vector machines have been used previously to price American options and what is introduced here is Bayesian techniques to both these approaches. Bayesian techniques used with both these approaches are compared in terms of pricing accuracy and time to train each of the learning algorithms. It was found that Bayesian SVM's out-performed Bayesian MLP's and that there is scope for further work. However, Bayesian SVM's took much longer to train than Bayesian MLP's even though they produced better error results.