關聯式分類是一種資料探勘方法，以關聯規則建構出分類系統。過去研究指出關聯式分類相較於傳統分類方法(如C4.5及ILP)，有較高的分類準確率，然而關聯式分類存在無法處理數值資料以及表達數值資料間關係之缺點。傳統分類方法中的歸納邏輯規劃 (ILP)具有易於關係表達以及對於問題表示與問題特定限制上較具彈性等優點。零容錯率、無法有效處理數值資料以及關係中的參數過多會影響處理效率是納邏輯規劃方法的缺點。本研究首先提出一個多層生物特徵結構的基因演算法(PGA)，改善歸納邏輯規劃系統的缺點。此結構可以表示數值資料間的關係，將之應用於關聯式分類規則編碼並建構出一個關聯式分類系統，以期兼具表達數值資料關係及高分類準確率之優點。實驗結果顯示本研究提出之方法(GA-ACR)具有高預測分類準確率，且優於根據資料分佈決定分類類別之資料分佈法。 Associative classification, one of data mining techniques, is a classification system based on associative classification rules. Although associative classification is more accurate than traditional classification approaches, such as C4.5 and ILP, it cannot handle numerical data and its relations. Therefore, an ongoing research problem is how to build associative classifiers from numerical data. Inductive logic programming (ILP), one of traditional classification approaches, has great capability of relations representation, and flexibility for problem representation and problem-specific constraints. However, it is not suitable for noisy environment and has weak facilities for processing numerical data, including unsatisfactory learning time with a large number of arguments in the relations. A phenotypic genetic algorithm(PGA) with multi-level phonotypic encoding structure is proposed to solve the problems in the ILP system. This structure has great capability of relations representation between numerical data and is used for relations encoding between numerical data in associative classification rules mining. The experiment results show that the proposed approach(GA-ACR) has high prediction accuracy and is highly competitive when compared with the data distribution method.
