本研究旨在設計一植基於模糊推論及模糊聚類法之模糊模式鑑別，以改善
傳統模糊模式鑑別法之缺點。本研究首先以模糊推論概念所推導的模糊模
式作為模糊模式鑑別的基礎，接著應用層級式模糊聚類法，將輸出資料分
類以決定模糊模式後件部的構造及前件部歸屬函數的初值，其次應用雙層
目標函數式模糊聚類法及模糊曲線鑑別模糊模式的重要輸入變數及模糊規
則數，最後，應用電腦模擬（MATLAB語言，4.0版），列舉幾個數值模擬
範例，以證實其效果。本研究之重要結論如下：一、藉由文獻探討知悉模
糊推論、模糊聚類的理論及模糊模式鑑別的方法，並能依據文獻的歸納選
擇適當的方法達成研究目的。二、本研究結合模糊推論及模糊聚類法之優
點，即應用層級式模糊聚類法依據輸入輸出資料的特性進行分類獲得前件
部初始參數，減少參數調整的時間，同時鑑別模糊模式後件部的參數及構
造。其次應用雙層目標函數式模糊聚類法及模糊曲線快速地鑑別模糊模式
的構造，減少模糊模式構造探索的時間及降低模糊模式構造鑑別的複雜性
，改善傳統模糊模式鑑別法之缺點。三、應用電腦模擬的結果證實層級式
模糊聚類法能將單輸入單輸出的系統，將不同類的資料連貫成為一個系統
；而雙層目標函數式模糊聚類法及模糊曲線對於多輸入單輸出的系統可快
速地鑑別出模糊模式的輸入變數，減少規則庫，節省推論的時間。因此，
本研究所提出之方法確實可行。

The purpose of this study was to design a fuzzy modeling
identification approach which was based on fuzzy reasoning and
fuzzy cluster and to improve the defects of traditional fuzzy
modeling identification approach. In this study, first applying
the fuzzy reasoning concepts induced the fuzzy model and made
the base of a fuzzy modeling identification. Then applying a
hierarchical fuzzy clustering approach to classify the output
data, it could identify the consequent structure of the fuzzy
model and set the initial value of membership function. Next, by
means of a double hierarchical objective function fuzzy
clustering identified the significant input variables and fuzzy
rules of fuzzy modeling. Eventually, this study tried to
illustrate some numerical examples to verify the validity of
those proposed method.The results of this study were as
follows:1. By reviewing literature understood the type of the
fuzzy reasoning and the fuzzy cluster and the fuzzy modeling
identification approach. According to the direction of this
study adopt proper method to achieve the purpose of this
study.2. This study applied a hierarchical fuzzy clustering
approach to classify the output data, and obtained the initial
value of membership function and reduced parameter tuning times.
Next in order to reduced times of a structure of fuzzy modeling
explored and reduced complexity of fuzzy model structure, and
improved defects of traditional fuzzy modeling identification
approach by a double hierarchical objective function fuzzy
clustering that identify significant input variables and fuzzy
rules of fuzzy modeling.3. Applying the results of computer
simulation verified that a hierarchical fuzzy clustering
approach could link different classification data of single-
input single-output system into one system. And a double
hierarchical objective function fuzzy clustering and fuzzy curve
could rapidly identify the input variables of fuzzy modeling for
the multi-input one-output system, and reduced rule base and
saved inference times. Therefore the results of this study was
verified useful.
The purpose of this study was to design a fuzzy modeling