Fast Eective Rule Induction

Many existing rule learning systems are computationally expensive on large noisy datasets In this paper we evaluate the recently proposed rule learning algorithm IREP on a large and diverse collection of benchmark problems We show that while IREP is extremely e cient it frequently gives error rates higher than those of C and C rules We then propose a num ber of modi cations resulting in an algo rithm RIPPERk that is very competitive with C rules with respect to error rates but much more e cient on large samples RIPPERk obtains error rates lower than or equivalent to C rules on of bench mark problems scales nearly linearly with the number of training examples and can e ciently process noisy datasets containing hundreds of thousands of examples