Learning If-Then Rules in Noisy Domains

Reference: P. Clark and T. Niblett. Learning if-then rules in noisy domains. In B. Phelps, editor, Interactions in Artificial Intelligence and Statistical Methods, pages 154-166. Gower, Hants, UK, 1987.

Abstract: This paper presents a description and empirical evaluation of a new induction system, CN2, which relaxes the requirement of complete consistency of rules with training data during rule generation. CN2 has been designed to induce short, simple, comprehensible rules in domains where problems of poor description language and/or noise may be present. Induced rules are in a form similar to production rules, with the condition being a conjunct of tests on an example's attributes and the conclusion being a class prediction. The paper can be viewed as an investigation of applying a method similar to tree pruning to the generation of production rule expressions.

Postscript: Not available. Email me with your postal address if you would like a hard-copy.