Accession Number : ADA292690
Title : Induction of Selective Bayesian Classifiers.
Descriptive Note : Technical rept.,
Corporate Author : INSTITUTE FOR THE STUDY OF LEARNING AND EXPERTISE PALO ALTO CA
Personal Author(s) : Langley, Pat ; Sage, Stephanie
PDF Url : ADA292690
Report Date : 15 AUG 1994
Pagination or Media Count : 11
Abstract : In this paper, we examine previous work on the naive Bayesian classifier and review its limitations, which include a sensitivity to correlated features. We respond to this problem by embedding the naive Bayesian induction scheme within an algorithm that carries out a greedy search through the space of features. We hypothesize that this approach will improve asymptotic accuracy in domains that involve correlated features without reducing the rate of learning in ones that do not. We report experimental results on six natural domains, including comparisons with decision-tree induction, that support these hypotheses. In closing, we discuss other approaches to extending naive Bayesian classifiers and outline some directions for future research. (AN)
Descriptors : *LEARNING MACHINES, *BAYES THEOREM, ALGORITHMS, EXPERIMENTAL DATA, NEURAL NETS, DECISION MAKING, PROBABILITY DISTRIBUTION FUNCTIONS, STATISTICAL INFERENCE, COMPARISON, ACCURACY, HYPOTHESES, CORRELATION TECHNIQUES, INDUCTION SYSTEMS.
Subject Categories : Statistics and Probability
Distribution Statement : APPROVED FOR PUBLIC RELEASE