自2001年以來，黃金價格受到新興市場成長等多重影響下不斷攀升，雖於2008年3月至2008年12月間進行了大幅修正，但近期仍不斷的創新高。加上台灣期貨交易所於2008年1月28日推出以台幣計價之黃金期貨，過去文獻鮮少以台幣黃金期貨作為研究標的，因此本研究以此作為研究對象，以發展出穩定的波段預測模型作為研究目的。

本研究樣本為2008/01/28至2010/12/31之台幣黃金期貨日資料，採用技術指標如：OSC_MA、OSC_KD及OSC_MACD，總體經濟資料如：台銀台幣黃金存摺賣出價、台指選擇權VIX指數、台美匯率、S&P GSCI指數、S&P 500 VIX指數、布蘭特原油及倫敦早盤黃金定價等資料作為輸入變數；並以隔日漲跌及隔日波段分別作為輸出變數共四種模型，以平滑支撐向量機進行模型訓練後比較，同時也探討加入停損策略前後對於報酬率的影響。

本研究發現：
一、 使用隔日波段作為輸出變數的準確率高於隔日漲跌，此能提供給波段操作投資人參考。
二、 輸入變數使用技術指標搭配總體經濟資料，並使用停損策略，能有較高的預測損益，且損益穩定向上成長，且統計上顯著優於不使用總體經濟資料的模型。
三、 使用波段預測進行交易，能避免多次進出造成交易成本過高，平均9.6個交易日(約兩周)才會進行反向交易

From year 2000, the gold price has been raised quickly and consistently by multiple reasons including the growth of emerging market. Although there was a drawdown between March 2008 and December 2008, the gold price continues to rise, with a breakout to $1500 recently. On January 28th, 2008 Taiwan Futures Exchange launched new futures contract on NT-Dollar gold price. Former studies seldom took it as their object for study; therefore we use this new derivative as our subject in order to develop a steady prediction model.

In this study, we employ a recent data between January 28th, 2008 and December 31th, 2010. Using technical indicators: OSC_MA, OSC_KD, OSC_MACD; and macroeconomics data: S&P 500 VIX, S&P GSCI, Taiwan VIX, Bank of Taiwan Gold Passbook price, London A.M Fix Gold Spot price, Brent Oil Price and the exchange rate between US Dollars and NT Dollars as our feature variables. We proposed four models based on different predict variables and feature variables, compared each other after model training with Smooth SVM and investigate the difference between with and without the stop-loss trading strategy.

Our findings are as follows:
1 Using next-day trend rather than next-day up/down as predict variable has higher accuracy, which could assist trend investors when making decisions.
2 Using macroeconomic data and technical indicators as feature variables and stop-loss trading strategy could have better accumulated payoff with steadily growth. Significantly better than models without macroeconomic data.
3 Using next-day trend as predict variable could avoid multiple trading; therefore reducing total trading cost. An average of 9.6 trading days will have different position.

中文部份
[1] 李映潔 (2007)，影響黃金價格因素其穩定性之研究，國立成功大學國際企業研究所碩士論文。
[2] 吳宗諭 (2010)，影響台銀黃金存摺價格因素之研究，國立中央大學產業經濟研究所碩士在職專班碩士論文。
[3] 余尚武、劉憶瑩、黃泓瑋 (2009)，動態避險模型之建立–以遠期外匯與一籃子貨幣避險策略為例，第15屆海峽兩岸資訊管理發展與策略研討會。
[4] 曹鈞斐 (2006)，運用平滑支撐向量機於期貨市場擇時策略之研究，國立台灣科技大學資訊管理系碩士論文
[5] 熊金志,胡金蓮,袁華強等 (2008)，光滑支持向量機的原理和進展，計算機工程第34卷第13期，pp172-173
[6] 趙智穎 (2010)，我國黃金期貨市場投資價值之分析，私立元智大學管理研究所碩士論文。
[7] 蕭翔薰 (2006)，遺傳演化類神經網路於黃金期貨價格預測之研究，私立真理大學財經研究所碩士論文。
[8] 闕彥菱 (2008)，利率、美元、黃金價格及原油價格之動態傳遞效果，國立高雄第一科技大學金融營運所碩士論文。
[9] 楊天立 (2009)，「第一次買黃金就賺錢」，高寶書版。
[10] 楊天立 (2006)，「黃金再起」，高寶書版。

 
英文部分
[1] Cohen, G., Qadan , M. (2010), “Is gold still a shelter to fear?”, American Journal of Social and Management Sciences, pp. 39-43
[2] Hsu, C. W, Change, C. C., and Lin, C. J. (2003), “A Practical Guide to Support Vector Classification”, Available at http://www.csie.ntu.edu.tw/~cjlin/papers/guide/guide.pdf
[3] Cao, L. J., Tay, F. E. H. (2003), “Support vector machine with adaptive parameters in financial time series forecasting”, IEEE Transactions on Neural Networks 14, No.6.
[4] Lee, Y. J., Mangasarian, O. L. (2001), “SSVM: a smooth support vector machine for classification”. Computational Optimization and Applications, 2001,22(1) : 5-21.
[5] Pai, P. F.; Wei, W. R. (2007), “Predicting movement directions of stock index futures by support vector models with data preprocessing”, Industrial Engineering and Engineering Management, IEEE International Conference , pp.169-173
[6] Lu , S. X., Wang, X. Z. (2004), “A comparison among four SVM classification methods: LSVM, NLSVM, SSVM and NSVM”, in: Proceedings of 2004 International Conference on Machine Learning and Cybernetics, vol. 7, pp. 4277–4282.
[7] Simpson, M. J., Svendsen, A., Chan, P. L. (2007), “Gold, Platinum, Silver. Demand and Supply in the International Finance Market: An Empirical Analysis”, City University of Hong Kong, Department of Economics and Finance.
[8] Sharpe, W.F. (1994), “The Sharpe ratio,” Journal of Portfolio Management 21 (1), pp. 49-58.
[9] Vapnik, V. N. (1995), “The Nature of Statistical Learning Theory”, New York: Springer.
[10] Huang , W., Nakamoria, Y., Wang, S.Y., “Forecasting stock market movement direction with support vector machine”,Computers & Operations Research, pp. 2513–2522
[11] Zhang , Y. J., Wei, Y. M. (2010), “The crude oil market and the gold market: Evidence for cointegration, causality and price discovery”, Resources Policy, Volume 35, Issue 3, September 2010, pp.168-177