Estimating Continuous Distributions in Bayesian Classi ers

When modeling a probability distribution with a Bayesian network, we are faced with the problem of how to handle continuous variables. Most previous work has either solved the problem by discretizing, or assumed that the data are generated by a single Gaussian. In this paper we abandon the normality assumption and instead use statistical methods for nonparametric density estimation. For a naive Bayesian classiier, we present experimental results on a variety of natural and ar-tiicial domains, comparing two methods of density estimation: assuming normality and modeling each conditional distribution with a single Gaussian; and using nonparamet-ric kernel density estimation. We observe large reductions in error on several natural and artiicial data sets, which suggests that kernel estimation is a useful tool for learning Bayesian models.