傳統的迴歸是假設觀測值的不確定性來自於隨機,模糊迴歸(fuzzy regression) 則是假設不確定性來自多重隸屬現象。模糊迴歸採用樣本模糊數(x , Y )來對模糊迴歸參 數進行估計,其中Y 為觀測模糊數。一般模糊參數A的估計方式是採用線性規劃,求出適當 的區間,來將觀測模糊數Y 的分佈範圍全部覆蓋住。我們認為此法不能真實地表達出樣本 所蘊含的資訊,本研究將另行設立一套模糊參數估計方法,此法對樣本的解釋方式將更為 合理,且估計的過程也比線性規劃簡便。迴歸常用來建構經濟和財務的模型,而此種模型 經常帶有模糊的特質,例如景氣循環、不規則趨勢等,本文將針對台灣景氣指標進行實務 分析,以此說明模糊迴歸模式的實用性。 In this paper, we propose a parameter estimation method for fuzzy regression models by using the fuzzy number and the method of least squares. Fuzzy regression models are frequently applied in economic or financial modeling. These models exhibit certain kind of linguistic requirements, such as the business cycle and the diversity trend. We take the linguistic prediction as our illustration example for demonstration. Empirical results demonstrate that our estimation procedure can determine fuzzy regression models effectively.