資產配置最適化是財務管理的重要議題，資產最適化的關鍵在於數學二次式的參數準確輸入，本論文的重點在於變異數的參數輸入。運用GARCH模型描述金融資產，並以樣本變異數法、CB拔靴法與PRR拔靴法估計波動度，達到降低投資組合的波動進而使資產配置最佳化，本論文研究結果顯示風險規避者建議採用PRR拔靴法，報酬追求者建議採用CB拔靴法。 Optimal assets allocation problem has become a critical issue in wealth management and it leads to the mathematical quadratic optimal problems. The key factors of making the “optimal” working are the input parameters’ accuracy. This thesis focuses on only one of the inputs, the variance estimation. By incorporating the well known GARCH model, the traditional sample variance estimation along with two other bootstrap variance estimation models, CB and PRR, are employed and applied to the optimal assets allocations problem in this research. Empirical results suggest that a risk averter should adopted CB bootstrap variance estimation method and a risk taker should adopted PRR bootstrap variance estimation method in constructing their optimal portfolios.