Attribute Selection Based on Reduction of Numerical Attributes During Discretization

Abstract

Some numerical attributes may be reduced during discretization. It happens when a discretized attribute has only one interval, i.e., the entire domain of a numerical attribute is mapped into a single interval. The problem is how such reduction of data sets affects the error rate measured by the C4.5 decision tree generation system using ten-fold cross-validation . Our experiments on 15 numerical data sets show that for a Dominant Attribute discretization method the error rate is significantly larger (5% significance level, two-tailed test ) for the reduced data sets. However, decision trees generated from the reduced data sets are significantly simpler than the decision trees generated from the original data sets.