使用關聯式法則解決決策樹不相關值問題

Remove The Irrelevant Values Problem in the Decision Tree by Using Association Rule

林靜宜(Ching-Yi Lin)

指導教授 ： 蔣定安

淡江大學/工學院/資訊工程學系碩士在職專班/碩士(2009年)

https://doi.org/10.6846/TKU.2009.00571

摘要

決策樹的特點是以遞迴的計算方式，透過由上而下和各別擊破的策略，但是在建構決策樹中有可能產生「不相關值問題」，同時因決策樹樹狀的結構，在移除不相關值後，切點的選擇也會因而改變，另一方面，決策樹在處理連續型數值屬性離散化是利用區域離散化的方式處理，而區域離散化必須考慮到屬性之間的相依性關係，若屬性之間不存在相依性關係，則全域離散化比區域離散化較好。
因此，在本文裡提出了移除不相關值問題的方法，並且利用關聯法則，整合分類法則樣版與全域離散化後的屬性切點，同時提出一個分類器建構方法以整合所有的分類法則。由本論文所提出方法的實驗結果，不僅移除了不相關值，且透過關聯法則整合全域切點後更精準。

關鍵字

決策樹 ； 不相關值問題 ； 離散化 ； 關聯式法則

並列摘要

The characteristics of the decision tree is based recursive formula, through top-down and divide-and-conquer. However, decision tree construction may have the irrelevant values problem, at the same time, the tree structure of decision tree, remove irrelevant values, the choice of cut-off point will also be changed. On the other hand, discretization techniques of the decision tree is local method , and local method must have between attributes and attribute dependency relationship, if there is no dependency, then the global of discretization methods are better than the local of discretization methods.

As a result, in this paper has been proposed to remove the irrelevant values problem, and the use of association rules, classification rules like the integration with the global discrete of the attributes after the cut-off point, at the same time to propose a method to build classifiers to integrate all of the classification rules. By the thesis of the proposed method of experimental results, not only to remove the relevant values, and using association rules integrate into the global cut-off point are more accurate .

並列關鍵字

Decision Tree ； Irrelevant Value ； Discretization ； Association Rule

參考文獻

[36]顏博文，「 應用資料探勘技術分析學生選課特性與學業表現」，中原大學，碩士論文，民國92年。

[37]Ding-An Chiang* , Cheng-Tzu Wang, Yi-Hsin Wang and Chun-Chi Chen,The Irrelevant Values Problem of Decision Tree for Improving a Glass Sputtering Process,2008.

[1]J.R.Quilan, Induction o decision trees, Machine Learning, 1, pp.81-106,1986.

[2]G.V.Kass, An exploratory technique for investigating large quantities of categorical data, Applied Statistics, vol.29, pp.119-127,1980.

[3]J.Cheng, U.M.Fayyad, K.B.Irani and Z.Qian,“Improved Decision Tree: a Generalized Version of ID3,“Proc. Of the Fifth Int’1 Conf. on Machine Larning,pp.100-108,1998.