AuthorsY. Wang, Q. Xin and F. Coenen
TitleHybrid Rule Ordering in Classification Association Rule Mining
StatusPublished
Publication TypeJournal Article
Year of Publication2008
JournalTransactions on Machine Learning and Data Mining
Volume1
Number1
Pagination1-15
Date PublishedJuly
Abstract

Classification Association Rule Mining (CARM) is an approach to classifier generation that builds an Association Rule Mining based classifier using Classification Association Rules (CARs). Regardless of which particular CARM algorithm is used, a similar set of CARs is always generated from data, and a classifier is usually presented as an ordered list of CARs, based on a selected rule ordering strategy. Hence to produce an accurate classifier, it is essential to develop a rational rule ordering mechanism. In the past decade, a number of rule ordering strategies have been introduced. Six major ones can be identified: Confidence Support & size-of-Antecedent (CSA), size-of-Antecedent Confidence & Support (ACS), Confidence Support size-of-Antecedent class-distribution-Frequency & Row-ordering (CSAFR), Weighted Relative Accuracy (WRA), Laplace Accuracy (LA), and Chi-square Testing (Łambda^2). Broadly speaking, these strategies can be categorized into two groups: Support-Confidence (including CSA, ACS and CSAFR) and Rule Weighting (including WRA, LA and Łambda^2). In this paper, we propose a hybrid rule ordering approach (framework) by combining one strategy taken from Support-Confidence and another strategy taken from Rule Weighting, which consequently develops nine rule ordering mechanisms. The experimental results demonstrate that all developed mechanisms perform well with respect to the accuracy of classification.

Citation KeySimula.ND.127