利用多層次類別優先度之規則排序以改善關聯式分類效能

Improving the performance of Associative Classification by using the Multi-level Class Priority of Rule Ranking

邱信淵(Hsin-Yuan Chiou)

指導教授 ： 蔣定安

淡江大學/工學院/資訊工程學系碩士班/碩士(2010年)

https://doi.org/10.6846/TKU.2010.00129

摘要

一般關聯式分類法(Associative Classification, AC)在規則排序(Ranking)[1][2]上，會先依照信賴度由高至低排序，接著依支援度由高至低排序，再依規則由短至長排序，短規則因為通用性較高，通常為了讓更多文件可以分類，因此短規則在排序上優於長規則，為了讓特殊文件能夠準確的分類，本論文採用了Lazy演算法，依照信賴度及支援度由高至低排序外，規則長度較長者先排序。
本論文核心為規則排列問題，除了採用Lazy法[3]所提出的排序法則為一般排序原則外，再加上本論文提出之多層次類別優先度來探討其對分類準確率及效能，並與Lazy演算法比較及對照。

關鍵字

關聯式法則 ； 多層次類別優先 ； 規則排序 ； 規則產生

並列摘要

In general, the approach in rule ranking of associative classification (AC)[1][2] begins first with confidence value in order of the highest to the lowest, then support value in order of the highest to the lowest, and finally rule in order of the shortest to the longest. In order to make more documents classifiable, short rules are ranked higher than long rules as short rules also have higher compatibility. With the use of discourse-based experiments in this study, it was found that common characteristics existed between certain categories and they were not always mutually associated. One could achieve a considerable degree of improvement by placing rules of a certain category in front of rules of another category.
The core of this paper is centered on the issue of rule ranking. Apart from adopting the ranking method proposed by Lazy[3] method as the general principle, Multi-Level class priority was proposed to explore its impact on the classification performance. It was proven in the experiments that adding Multi-Level class priority in rule ranking would help to achieve better classification performance than any general ranking principles.

並列關鍵字

Associative Rule ； Multi-level class priority ； Rule Ranking ； Rule Generation

參考文獻

[2]F. Thabtah. , ”A review of associative classification mining.” Knowl. Eng. Rev. , vol. 22, 2007 pp. 37-65.

[3]E. Baralis and P. Garza. ”A lazy approach to pruning classification rules.” Data Mining, IEEE International Conference on 0pp. 35. Dec. 2002.

[5]Han-Sheng Hsiung.”Improving the accuracy of text categorization by using association rule with class priority.” Master thesis of Tamkang University, Jun. 2009, pp. 1-54.

[6]M. L. Antonie and O. R. Zaiane. ”Text document categorization by term association.” Proceedings of the 2002 IEEE International Conference on Data Mining.

[9] W. Li, J. Han, and J. Pei, CMAR: Accurate and efficient classification based on multiple class-association rules., ICDM-01, San Jose, CA, Nov. 2001, pp. 369-376.