Publication Abstract

Dain, Oliver, Boyer, Stephen, and Cunningham, R. K. IREP++, a Faster Rule Learning Algorithm. Proceedings of the 2004 SIAM International Conference on Data Mining.

Abstract

We present IREP++, a rule learning algorithm similar to RIPPER and IREP. Like these other algorithms IREP++ produces accurate, human readable rules from noisy data sets. However, IREP++ is able to produce such rule sets more quickly and can often express the target concept with fewer rules and fewer literals per rule resulting in a concept description that is easier for humans to understand. The new algorithm is fast enough for interactive training with very large data sets.

 

top of page