Work

BayesClass: an R package for learning Bayesian network classifiers

Year: 2013
Type: conference-paper
Abstract: BayesClass implements ten algorithms for learning Bayesian network classifiers from discrete data. This includes score+search algorithms and those that maintain the structure of a naive Bayes but extend it with additional parameters. Many of the algorithms perform implicit feature selection. Most of the structure searching methods are based on the forward search heuristic while score functions include classifier accuracy, likelihood, and a function based on the significance of dependency between two sets of variables (see Blanco et al. (2005)). Implemented algorithms include: tree augmented naive Bayes (TAN) (Friedman et al. (1997)), forward sequential selection (Langley and Sage (1994)), forward sequential selection and joining (FSSJ) (Pazzani (1996)), and adaptations of those methods from (Blanco et al. (2005)); the adjusted probability naive Bayesian classification (Webb and Pazzani (1998)), the attribute weighted naive Bayes (Hall(2007)), and others. We also propose an adaptation of the TAN algorithm to operate on structures learned by FSSJ.
Source: Hispana
Authors: Mihaljevic, Bojan, Larrañaga Múgica, Pedro María, Bielza Lozoya, María Concepción
Language: English
Open Access status: green
OpenAlex ID: W7055002865
FWCI:
Cites: 5
Cited by:
Related to: 9