Fuzzy Bayes and Fuzzy Markov Predictors

Issue title: SBRN'02

Article type: Research Article

Authors: Teixeira, Marcelo Andrade^; | Zaverucha, Gerson

Affiliations: Systems Engineering and Computer Science – COPPE, Federal University of Rio de Janeiro (UFRJ), Caixa Postal 68511, 21945-970, Rio de Janeiro, Brazil. Tel.: +55 21 2562 8675; Fax: +55 21 2562 8676; E-mail: [email protected] | Electric Power Research Center (CEPEL), Rio de Janeiro, Brazil

Note: [] Corresponding author

Abstract: Two fuzzy probabilistic predictors, the Fuzzy Bayes Predictor (FBP) and the Fuzzy Markov Predictor (FMP), are described and evaluated for the task of single-step monthly electric load forecasting. They are modifications of the Naive Bayes Classifier and Hidden Markov Model (where temporal dependence among the examples is ignored and taken into account, respectively) to enable them to predict numerical values by incorporating features from fuzzy systems. This is a generalization of the regression-by-discretization approach of using classifiers for regression. These hybrid systems are successfully compared with two traditional forecasting methods, Box-Jenkins and Winters exponential smoothing.

Journal: Journal of Intelligent & Fuzzy Systems, vol. 13, no. 2-4, pp. 155-165, 2002/2003