傳統上，時間數列的型態分類與認定的分析方法，一般都應用在定態的隨機過程。雖然單根檢定的方法用來檢定一時間數列是否定態的判定，一直被計量經濟學家所重視。但由於近幾年來非直線性時間數列越來越受到重視與研究，以傳統的單根檢定法來分析已無法顯出其數列的特性，甚至許多時候會導致對其數列辨識準確度穩健性的喪失，所以結構轉變的檢定先行於單根檢定，對於非線型時間數列來說是非常重要的。

　　鑑於此，本文嘗試模擬5組非線性的時間數列資料，用平均值估計信賴區間之方法和模糊熵分類方法，以事前的觀點用我們所定義的分類標準客觀的來認定其結構轉變的區間。我們也舉出匯率的實際例子以我們所提出的方法加以探討認定，找出其高低匯率時期及轉型期。

　　Traditionally, the analysis methods of pattern classification and recognition for time series generally apply to the stationary process. Tests for unit roots used to test whether the time series is stationary has always been looked upon by the statisticians econometrics. Because there have been much more research on the nonlinear time series in recent years, the tests for unit roots can't tell the features of time series and even result in the lost of robustness for the identification precision of the time series.

　　Because of this fact, this article tries to simulate four sets of nonlinear time series data, using both the method for estimating confidence interval based on the mean and the one for fuzzy classification to recognize the structure change period objectively with the classification standards we define in the perspective view. We also take the exchange rate as an example and recognize it using method to find out the high and low exchange period and it's structure change period.