本文利用2003年1月1日至2003年12月31日間在台灣期貨交易所發行之台灣股票選擇權,應用Neuro-Fuzzy評價模型,分別搭配四種波動率模型,歷史波動率、隱含波動率、GARCH波動率、GJR-GARCH波動率,對台灣七檔股票選擇權進行評價與套利。評價績效之指標為平均絕對誤差(MAE)、平均絕對誤差率(MAPE)、誤差均方根(RMSE)與Theil's U 進行評估,並且利用Mann-Whitney-Wilcoxon檢定。 研究結果顯示,在搭配不同波動率下之Neuro-Fuzzy評價模型的績效,以歷史波動率、GARCH波動率、GJR-GARCH波動率評價績效最好。在套利方面,分別對七檔股票選擇權進行套利交易模擬,研究結果發現,有五檔股票選擇權可獲得超額報酬。
The study is based on the transaction data from the Taiwan stock option issued by TIFEX from January 1, 2003 to December 31, 2004. Applying Neuro-Fuzzy with four volatility models, namely historical volatility, implied volatility, GARCH volatility, GJR-GARCH volatility to the pricing and arbitraging of seven stock options. Further the study evaluates the efficiency of pricing performance of Neuro-Fuzzy models on the prices of the stock option. The MAE, MAPE, MSE, Theil's U and Mann-Whitney-Wilcoxon test are applied to evaluate the pricing performance. Empirical results show that pricing performance of historical volatility, GARCH volatility, GJR-GARCH volatility is better than implied volatility models. Among seven stock options return, we find that the five obtain excess return.