從關聯式資料庫系統產生乏晰規則以作乏晰資訊擷取的新方法

Abstract

本論文提出了一個從關聯式資料庫中建構出乏晰決策樹，並從而產生乏晰 規則的乏晰觀念學習系統演算法，文中並對所建構出的乏晰決策樹之完整 性也一併討論。 根據所產生之乏晰規則，我們也提出了一個方法以預測 關聯式資料庫中的未知值。 在本論文中，我們也作了一個實數值函數的 近似能力的分析實驗，將我們所提出的乏晰觀念學習演算法與現存的方法 來作一比較。 從實驗結果得知，本文中所提之乏晰觀念學習演算法的整 體結果較現存的方法佳，特別是函數為f(x) = x/2時。 另外，在本論文 中也提出一個新的資料分群演算法以作資料庫系統之乏晰查詢處理。此方 法較現存的方法更具彈性及更有效率， 此乃因此演算法具有下列優點： (1) 群數不必預先設定。 (2) 可動態地更改乏晰詞的範圍。(3) 不必作 複雜的歸屬函數計算。(4) 查詢速度較快。 In this thesis, we present a fuzzy concept learning system algorithm (FCLS) to construct fuzzy decision trees from relational database systems and to generate fuzzy rules from the constructed fuzzy decision trees. The completeness of the constructed fuzzy decision tree is alsodiscussed in details. Based on the generated fuzzy rules, we also present a method to forecast null values in relational database systems. Furthermore,we also made an experiment to compare the proposed FCLS algorithm with the existing methods for analyzing the ability of approximation ofreal-valued functions. The experiment result shows that the overall result of approximation of the FCLS algorithm is better than the existing methods,especially when f(x)=x/2. Furthermore, we also present a new clustering algorithm to deal with fuzzy query processing for database systems. Theproposed algorithm is more flexible and more efficient than the existing method due to the fact that the proposed algorithm has the following good features: (1) The number of clusters does not need to be predefined. (2) The ranges of fuzzy terms can dynamically be changed. (3) It does not need to perform complicated membership function calculations. (4) The speed of fuzzy query processing can be much faster.