Ringnorm dataset

This is an implementation of Leo Breiman's ringnorm example[1]. It is a 20 dimensional, 2 class classification example. Each class is drawn from a multivariate normal distribution. Class 1 has mean zero and covariance 4 times the identity. Class 2 has mean (a,a,..a) and unit covariance. a = 2/sqrt(20). Breiman reports the theoretical expected misclassification rate as 1.3%. He used 300 training examples with CART and found an error of 21.4%.

Dataset profile:

Origin: artificial

Usage: historical

Number of attributes: 21

Number of cases: 7,400

Number of prototasks: 1

Number of methods run on this dataset: 3

Download ringnorm.tar.gz

Contributed by: Michael Revow.



(1) Breiman L. Bias, variance and arcing classifiers. Tec. Report 460, Statistics department. University of california. April 1996.


Last Updated 11 October 1996
Comments and questions to: delve@cs.toronto.edu
Copyright