| Summary: | 碩士 === 中華大學 === 資訊管理學系(所) === 97 === In September, 1997, Taiwan issued warrant in the derivative financial market for the first time, and then the newly developed market of financial derivatives became flourished in the recent years. Warrant is a kind of right agreement that after the investor pays amount of cost to get the warrant, he can buy underlying asset share from the distributor with the strike price within the expiration date, or he can collect the price difference by cash payment.
We apply the Black-Scholes pricing method, Grey model, Cumulate 3-point Least Square Prediction Model (C3LSP) with genetic algorithm (GA) based back-propagation neural network (BPN) to explore the dynamism of warrant. Black-Scholes pricing method is used to judge the reasonableness of price and to decrease predict error. The Grey model is used to decrease the influence degree of data noise on the predict model. C3LSP is used to improve the overshoot phenomenon of residual error produced by grey model. Then, genetic algorithm is applied to select the input variables and parameter settings for SVM. The data of input variables is used to train SVM. Finally, the trained SVM is used to testify the testing data to find better performance.
In empirical results, the proposed method can improve classification effectiveness. And, there is apparent behavior to the improvement of efficiency to join GM-C3LSP.
|
|---|
