SVM Evaluates Index Options Under Nonlinear GARCH Volatility Estimates

• Main Authors: Li Jui-Kai, 李瑞凱
• Other Authors: 盧智強
• Format: Others
• Language: zh-TW
• Published: 2007
• Online Access: http://ndltd.ncl.edu.tw/handle/25128267283163937794

碩士 === 國立臺北商業技術學院 === 商學研究所 === 95 === It can’t be explained the problems of leptokurtic and clustering under the BS model because of the assumption of constant volatility. In order to overcome the difficulties, scholars used linear and nonlinear GARCH models to estimate volatility. However,there were still no consistent results in the past research regarding to pricing on different options under different models.This study attempt to estimate the volatility of TEO under linear GARCH and nonlinear GARCH models.The result of estimation shows that TBGARCH is the minial value in average among all and the plot tend to more flat day by day.For GJRGARCH,the value of volatility is a bit bigger than TBGARCH,but still in good situation.Further,we use the volatility to be one of input variable to evaluate the option price under Support Vector Machine model(SVM) and Back-propagation Neural(BPN) Networks .To compare the performance among three models,SVM,BPN,and BS.The result shows SVM is more accurate than other two models in pricing TEO.The value of MAPE or RMSE is better when using the volatility under TBGARCH,EGARCH,GJRGARCH models estimation to be the input volatility.The main purpose verifies that SVM is the better pricing model than BPN.