股票是一般投資大眾最熟悉也最慣常使用之投資標的，股票選擇權則較少與人所熟知。本研究提供了股票選擇權避險策略，讓投資大眾能夠理解若適當應用股票選擇權，則在多、空頭市場中皆能達到有效避險與穩定收益。 研究中將使用三種避險策略，分別為買入持有、買入並全面避險以及買入並避險擇時策略。買入並避險擇時策略應用傳統時間序列GARCH模型、人工智慧SSVR、GASVR模型，預測未來股價波動走向，建立相對應之避險部位。 透過實證研究結論，本研究所提出之避險策略，能夠在投資股票時，依照未來走勢的可能性建立避險部位，避險策略結果說明，股票選擇權的確為投資人可用來避險的衍生性金融商品，以及避險策略在平衡投資績效上的必要性。 本研究樣本為聯電、國泰金、中信金與友達個股，探討時間序列模型與人工智慧模型的預測能力，並以聯電、國泰金、中信金與友達個股選擇權為標的進行避險策略，進行報酬率與避險績效之評估。本研究實證結果如下：
一、預測股價方面，人工智慧模型於預測誤差與方向上優於時間序列模型。
二、在買入並全面避險策略中，期末損益為虧損居多，虧損機率達75%，但對於波動幅度大之股票標的具有投資效益。在買入並避險擇時策略中，期末損益盈餘機率達到75%，若不計時間序列模型，則人工智慧模型獲利機率達到100%，此外，人工智慧模型擇時建立避險部位的勝率普遍達到70%以上。
三、以避險績效來說，全面避險與避險擇時策略皆具能有避險功能，在避險擇時策略當中，人工智慧模型的避險績效經檢定證實較時間序列模型更佳。

Stock is the most familiar investment tool to investors, and people put it in the first place to use. But stock option isn’t like stock; so many investors even don’t know how stock option works. In this research, stock option hedging strategy is mentioned to investors, so that investors can realize that no matter in bull or bear market, the hedging strategy always can reduce the risk of investment and stable the return. In this research, three kinds of hedging strategies: include no hedge, all covered hedge, and timing chosen hedging strategy are used, and the timing chosen hedging strategy use GARCH, SSVR and GASVR as prediction models to predict stock behavior and build the hedging position. Through the experiment, the hedging strategies in this research can build the right hedging position with referring the prediction of future stock. The experiment result is about stock option can be used into hedging strategy, and it’s necessary to use a hedging strategy to balance the performance of investment. Three Samples, UMC, Cathay, Chinatrust and AUO stocks are tested for capacity of the model’s prediction. Those stock options are applied into hedging strategy and the result of hedging strategy is evaluated. The empirical results are as follow:
1. In predicting stock prices, artificial intelligence model in prediction error and aspect of prediction is generally superior to time-series model.
2. In all covered hedging strategy, it’s usually in the loss situation, the profit loss rate reach 75%. But in the timing chosen hedging strategies, the winning percentage can reaches 75%. If we ignore the time series model, the winning percentage of artificial intelligence model even reaches 100%
3. In the hedging performance, all covered hedging strategy and timing chosen hedging strategy both have positive hedging capabilities, moreover, through the experiment, the hedging performance of artificial intelligence model is better than the time series model.

[1]. 邱韻蓉(2009)，「應用平滑支撐向量迴歸與灰預測於臺灣加權股價指數真實波動率之研究」，臺灣科技大學資訊管理研究所碩士論文。
[2]. 林大仁(2007)，「運用支撐向量迴歸與倒傳遞類神類神經網路建構產品造形設計之預測模式」，國立成功大學工業設計學系碩士論文。
[3]. 周資凱(2006)，「基因演算法應用在選擇權避險策略之研究」，國立高雄第一科技大學風險管理與保險所碩士論文。
[4]. 陳寬裕(2006)，「結合遺傳演算法與支援向量迴歸於臺灣股票加權指數之預測」，計量管理期刊，第三卷，第一期。
[5]. 張愉佳(2006)，「財務報酬波動之預測：靴帶抽樣方法與應用」，國立政治大學經濟系碩士論文。
[6]. 黃敏菁(2005)，「支援向量機在財務時間序列預測之應用」，天主教輔仁大學金融研究所碩士論文。
[7]. 黃祺偉(2009)，「多核心支援迴歸向量機應用於股價預測」，國立中山大學電機工程學系碩士論文。
[8]. 鄭亦妏(2003)，「在Black-Scholes 評價模型下台指選擇權最適波動性估計方法之研究」，淡江大學管理科學學系碩士論文。
[9]. 劉洪鈞、李命志、賴曉萍(2009)，「臺灣50ETF與台指期貨避險績效之評估」，金融創新與科技整合學術研討會，民國九十八年六月十一日。
[10]. 劉哲良(2008)，「期貨商品最適預測模型之研究－以臺灣市場為例」，臺灣大學國際企業學研究所碩士論文。
[11]. 蔣輝、王志忠(2009)，「基於灰色系統的支援向量迴歸預測方法」，湖南大學經濟數學期刊，第26 卷，第2期。
59
[12]. 薛立言、賴靖宜(2007)，「一般化期貨與選擇權之交易策略研究」，中華民國期貨業商業同業公會委託研究專案。
[13]. 臺灣期貨交易所 (2002)編印之「股票選擇權規劃書草案」。
[14]. Akgiray, V. (1989), “Conditional Heteroscedasticity in Time Series ofStock Return: Evidence and Forecasts”,Journal of Business, (62), pp.55-80.
[15]. A. J. Smola, Ph. D. Dissertation (1998), “Learning with kernels”,Technische Universitat Berlin.
[16]. Bollerslev, T., Chou, Y. R., & Kroner, F. K. (1992), “ARCH modeling in finance”. Journal of Econometrics, 52, 5-59.
[17]. Bollerslev, T. (1986),“Generalized Autoregressive Conditional Heteroskedasticity”,Journal of Econometrics, Vol.31, pp.307-327.
[18]. D.E. Goldberg (1989), Genetic Algorithms in Search, Optimization and Machine Learning, Addison Wesley.
[19]. Engle, R. F. (1982),“Autoregressive Conditional Heteroskedasticity with Estimates of The Variance of UK Inflation”, Econometrica, Vol.50, pp.987-1008.
[20]. Holland, J. (1975). “Adaptation In Natural and Artificial Systems.“, The University of Michigan Press, Ann Arbour.
[21]. Hsu, C.-W., Chang, C.-C. and Lin, C.-J. (2003),“A Practical Guide to SupportVector Classification. Technical Report”, Department of Computer Science andInformation Engineering, National Taiwan University.
[22]. Lee, Y.-J., Hsieh, W.-F., Huang, C.-M. (2005),“ε-SSVR: A smooth supportvector machine for ε-insensitive regression“, IEEE Transactions on Knowledgeand Data Engineering, Vol.17(5), pp.678-685.
[23]. Ljung, G. M. and Box, G. E. P. (1978),“On a measure of lack of fit in time series models. “,Biometrika 65, 297–303.
60
[24]. Park, T. H. and L. N. Switzer (1995), “Bivariate GARCH Estimation of the Optimal Hedge Ratios for Stock Index Future: A Note“, Journal of Futures Markets, 15, P. 61-67.
[25]. Said &Dickey(1984) ,“Testing for Unit Roots in Autoregressive-. Moving Average Models of Unknown Order.“, Biometrika, 71, 599-607.