Research, Artificial Neural Network-based Options Pricing

Warrant market is a highly complicate nonlinear system. Its variation has its own regulation, but also is influenced by many other factors such as market, economy and non-economy. It will be very helpful to investors, if we could set up an objective scientific warrant pricing method.Artificial neural network (ANN) is a rising borderline science. Compared to the traditional warrants pricing methods (B-S model) , it doesn't need exact mathematical model and does not have any precondition or assumption. It can make up the deficiency of traditional pricing methods and solve some problem that traditional pricing methods failed to resolve.The paper uses the BP artificial neural network and Genetic Algorithms to establish a warrant pricing model, takes a sample of 24 European Warrants now existed and listed before January 1, 2007. The analysis period is the 30 consecutive trading days after listed, the data of first 20 days are used as study samples and the remaining are used as testing samples. The volatility of target stock is an important variable that affects the value of warrants. The paper estimates the historical volatility and the implied volatility of 14 warrants, takes them as the input variable of BP model separately to compare their impact.Comparing BP model and B-S model shows that, no matter using the fluctuating rate of history or the fluctuating rate of imply, the BP model is superior to the B-S model in the pricing precision, and after using the fluctuating rate of imply, the BP model reduce the deviation with market price in the whole.