本文應用Markowitz(1952) 的平均數-變異數模型(Mean-Variance Model)來探討最適資產的配適。實證上選取台灣五十指數成分股為投資組合標的,經由動態的移動視窗法,以2008年9月5日至2009年8月28日共51筆週報酬資料做為第一期樣本資料,利用動態的方式對不同的模型去估計投資組合成分股下一期平均數、變異數和共變異數,再藉由平均數-變異數模型,導出下一期所應投入的投資組合最適權重。根據這些權重來預測2009年9月4日至2010年8月27日所應該要配適的最佳投資組合。透過不同的績效指標進行評估,發現利用DCC-GARCH模型去估計參數,相較於傳統平均數-變異數模型以歷史資料評估不但較能正確反映市場走勢,也有較佳的預測能力及投資績效。 This paper use the Mean-Variance Model of Markowitz(1952) to discuss the allocation of the optimal assets. We use the component stocks of Taiwan 50 Index for the underlying of the portfolio. Through the dynamic moving window estimation, we use the weekly return from 2008/9/5 to 2009/8/28 for the first sample data. Estimate the portfolio’s mean, variance and the covariance of the next period of different kinds of model by dynamic method, then using the Mean-Variance Model to derive the optimal weight of the portfolio. We use this weights to forecast the allocation of the optimal assets from 2008/9/4 to 2010/8/27. This paper found that through the estimation of the different performance index, compare to the Mean-Variance Model, DCC-GARCH model is not only more accurate to reflect the market trend but also has better forecast ability and investment performance.