The Application Of RBF Neural Network In The Pricing Of Warrants

Warrants are actively traded in China' s capital market currently.However,according to empirical evidence,there exists a huge gap between model price and market price.Hence warrant pricing is becoming an important issue demanding further studies.Warrant has the functions of price-detection and risk-evasion,while it also has great risk of itself.Therefore,warrant should be treated with rational attitude and its real price should be fixed according to scientific analysis.An accurate and scientific method has not been found for the pricing of warrants.Warrant is actually a kind of option,it is usually valued by the classic option pricing model.However,the assumption assumed by the model cannot be satisfied under our market environment,so the model result deviates from the market price.The neural network has become a useful tool in the financial decision of the western developed country. but it just starts to be developed in our country.This paper used RBF neural network to price seven warrants of different duration,based on the analysis of the Black-Scholes pricing model.According to the error analysis of the pricing result,the paper concludes that the Black-Scholes model has important value on the price prediction of the warrant and RBF neural network model has the better pricing result than the Black-Scholes model.However,RBF neural network has several flaws.Firstly,the construction of the RBF neural network model must be based on the history transaction data Secondly,the application needs to do miscellaneous work.The third is the pricing result goes worse when the history data fluctuates.