近年來世界金融環境快速發展,銀行業務廣泛而複雜,導致控管作業風險的重要性越來越高。由於作業風險資料的厚尾特性,本研究除了以極值理論(Extreme-value Theory)模型來評估其風險值外。更進一步嘗試結合GARCH模型與極值理論模型的方法評估其風險值,除了抓住資料尾端特性之外,也希望能掌握風險值隨著時間變動的特性。本文實證研究期間為1995年至2009年,共19年期間台灣商業銀行的作業風險資料。實證結果發現: GARCH的極值理論模型能更精確的抓住作業風險隨時間變化的特性,充分掌握損失的變化,達到更精確的效果。 Because of the fast-developing of finance environment recently, the importance of risk management in banking business has been heightened. This article takes several commercial banks in Taiwan during the period from 1995 to 2009 as example, analyses the tails’ characteristics of operational risk loss event. It measures the fat-tail loss by the POT model of EVT. We further use GARCH-EVT model to capture time varying feature of data, so that we can better understand the characteristic of operational risk.