Feature extraction with hybrid neural networks

Neural networks (NN) and fuzzy logic systems (FLS) are used successfully for financial forecasting, credit rating and portfolio management. In search for more sophisticated modeling techniques a mixture of NN and FLS has proved to be worth consideration. We propose the novel constructive approach by which a neuro fuzzy network is built up with the help of a constrained optimizer. The mathematical motivation for such hybrid networks is presented, using the Kolmogorov theory of metric entropy. As an application of the proposed approach we build a neuro fuzzy network model which is able to explain the prices of call options written on the S&P 500 stock index. While option pricing theory typically requires a highly complex statistical model to capture the empirical pricing mechanism, our results indicate that this algorithm leads to more parsimonious functional specificationes which have a superior out-of-sample performance.