New Insights on Nontechnical Losses Characterization Through Evolutionary-Based Feature Selection
Nenhuma Miniatura disponível
Data
2012-01-01
Autores
Oba Ramos, Caio Cesar
de Souza, Andre Nunes
Falcao, Alexandre Xavier
Papa, João Paulo [UNESP]
Título da Revista
ISSN da Revista
Título de Volume
Editor
Institute of Electrical and Electronics Engineers (IEEE)
Resumo
Although nontechnical losses automatic identification has been massively studied, the problem of selecting the most representative features in order to boost the identification accuracy and to characterize possible illegal consumers has not attracted much attention in this context. In this paper, we focus on this problem by reviewing three evolutionary-based techniques for feature selection, and we also introduce one of them in this context. The results demonstrated that selecting the most representative features can improve a lot of the classification accuracy of possible frauds in datasets composed by industrial and commercial profiles.
Descrição
Palavras-chave
Feature selection, gravitational search algorithm, harmony search, nontechnical losses, optimum-path forest, particle swarm optimization, pattern recognition
Como citar
IEEE Transactions on Power Delivery. Piscataway: IEEE-Inst Electrical Electronics Engineers Inc, v. 27, n. 1, p. 140-146, 2012.