本研究根據國際貨幣基金組織(IMF)外匯儲備貨幣(COFER)的七種貨幣加上人民幣共八種幣別為投資組合的實證對象。利用Markowitz提出的MV模型建立最適的外幣投資組合並與等權重模型來評估投資績效,其次,再分別以變異數-共變異數模型及DCC-GARCH模型來衡量風險值,最後進行回溯測試探討不同的投資組合模型,是否可以真實的反映出面臨的市場風險。研究結果發現,變異-共變異數模型(VAR-COV)風險值較DCC-GARCH風險值模型為佳,且經由用MV模型建構的投資組合方式可降低且分散投資人的風險,達到風險程度最小的目的。 This study take seven International Monetary Fund (IMF) reserve currency (COFER) currency and Renminbi as a foreign portfolio. We firsy apply Markowitz(1952)''s MV model to establish the optimal currency portfolio. In addition, we compare its performance with an equal weighted protfolio model. Secondly, We use the Variance-Covariance model and DCC-GARCH model to measure the portfolio Value at Risk, and finally using the back-testing to test the performace of the two models. The results show that, DCC-GARCH-VaR is better than the VAR-COV-VaR model, and through a portfolio approach with MV model construction can reduce the investor''s risk diversification and achieve the purpose of the minimum risk.