2010 | OriginalPaper | Buchkapitel
Analyzing the Impact of the Discretization Method When Comparing Bayesian Classifiers
verfasst von : M. Julia Flores, José A. Gámez, Ana M. Martínez, José M. Puerta
Erschienen in: Trends in Applied Intelligent Systems
Verlag: Springer Berlin Heidelberg
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by
Most of the methods designed within the framework of Bayesian networks (BNs) assume that the involved variables are of discrete nature, but this assumption rarely holds in real problems. The Bayesian classifier AODE (Aggregating One-Dependence Estimators) e.g. can only work directly with discrete variables. The HAODE (from Hybrid AODE) classifier is proposed as an appealing alternative to AODE which is less affected by the discretization process. In this paper, we study if this behavior holds when applying different discretization methods. More importantly, we include other Bayesian classifiers in the comparison to find out to what extent the type of discretization affects their results in terms of accuracy and bias-variance discretization. If the type of discretization applied is not decisive, then future experiments can be
k
times faster,
k
being the number of discretization methods considered.