A Supervised Discretization Algorithm For Multivariates

Suppose we are given a data set with a categorical response variable and high dimensional continuous or categorical explanatory variables.In customer relation management practice,an ideal executable profile usually consists of categorical or discretized variables.They are readily identified customer characters,which portrayed classified data forms.This requires us to discretize continuous vari-ables.There are two types of discretization of continuous vaviables as covariates:supervised discretization and unsupervised discretization.A proper supervised dis-cretization usually achieves a better profile than the unsuprevised ones do.Huang-Pan-Wu et al had verified the reliability of the conclusions.For high dimensional data,a forward supervised discretization algorithm is proposed to capture higher(given)association with the response variable than that of the independent(or individual)supervised discretization algorithm recently proposed by Huang-Pan-Wu.The Goodman-Kruskal ? and ? based discretization experiments are also presented.