近年來,高頻金融資料和超高頻金融資料的研究已經成為一個備受矚目的研究領域,會成為如此熱門的研究,在於高頻金融資料和超高頻金融資料本身所富含的訊息,隨著高頻資料時代的來臨,人們所能夠掌握的資訊也越來越複雜,讓本來就瞬息萬變的金融市場更加難以捉摸,要如何在這些豐富的訊息中來進行分析和探討,便成為了近年來非常熱門的一個研究領域。本研究主要為探討使用超高頻資料估計變異數並代入平均數-變異數投資組合模型,進行資產配置最佳化,透過觀察累積報酬率的變化,找出在均異效率投資組合下,使用超高頻資料來估計變異數,會如何對投資績效產生影響,並歸納出其特性。主要是透過比較不同公司持股數下採用的三種報酬率估計模型(樣本平均數、資本資產定價模型、三因子模型)與三種變異數估計模型(樣本變異數、指數加權移動平均、已實現波動度)搭配產生的投資組合的投資績效比較來進行分析。實證分析的結果顯示,將超高頻資料使用已實現波動度來估計變異數並代入平均數-變異數投資組合模型進行資產配置最佳化,所得到的投資組合,能夠快速的反映市場狀況,在多頭市場下短時間內能夠快速的獲利,適合積極型的投資人使用此投資策略。 In recent years, the related research of high-frequency and ultra high-frequency financial data has drawn a lot of attention since its abundant information. In high-frequency data era, people are able to master more complicated information, analyze and explore the sufficient information. Therefore, high-frequency financial data has become a popular research field.This study investigates the incorporation of ultra-high frequency data with variance estimation in efficient mean-variance portfolio model for optimal assets allocation problem. Three mean estimation models (sample mean, the capital asset pricing model, three-factor model) and three variance estimation models (sample variance, exponentially weighted moving average, realized volatility) are used as the input in our optimal assets allocation problems. It identifies how the use of ultra-high frequency data incorporates variance estimation will impact on investment performance in terms of cumulative rate of returns. Different portfolio size effects are also studied. Empirical results shows that the portfolio uses ultra-high frequency data to estimate variance through Andersen’s realized volatility model for optimal asset allocation problems can quickly reflect market’s fluctuation in the next bull market within a short period and lead to profit gain.