Rule Induction with Extension Matrices

This paper presents a heuristic, attribute-based, noise-tolerant data mining program, HCV (Version 2.0), based on the newly-developed extension matrix approach. By dividing the positive examples (PE) of a speciic class in a given example set into intersecting groups and adopting a set of strategies to nd a heuristic conjunctive formula in each group which covers all the group's positive examples and none of the negative examples (NE), the HCV induction algorithm adopted in the HCV (Version 2.0) software nds a description formula in the form of variable-valued logic for PE against NE in low-order polynomial time at induction time. In addition to the HCV induction algorithm, this paper also outlines some of the techniques for noise handling and discretization of numerical domains developed and implemented in the HCV (Version 2.0) software, and provides a performance comparison of HCV (Version 2.0) with other data mining algorithms ID3, C4.5, C4.5rules and NewID in noisy and continuous domains. The empirical comparison shows that the rules generated by HCV (Version 2.0) are more compact than the decision trees or rules produced by ID3-like algorithms, and HCV's predicative accuracy is competitive with ID3-like algorithms.