Título: Sistema Híbrido Baseado Em Extreme Learning Machine E Harmony Search
Autores: Almeida, Edgar Wellington Marques de; Valença, Mêuser Jorge da Silva
Resumo: In this work was used a technique of seek associated to a training algorithm of artificial neural network. The aim is to develop a hybrid system which can be used for people from others knowledge area. Besides generate the weight matrix, it selects the input variables and builds the neural network architecture, choosing the number of neurons in the hidden layer. The neural network evaluating metrics guide the search, choosing the configurations which show the minor: forecast error, input variables and neurons in the hidden layer. In general, the hybrid system reached a satisfactory performance, obtaining a better or equivalent response to the methods used by National Operator of Electrical System (ONS) .
Palavras-chave: Artificial Neural Network; Variables Selection; Extreme Learning Machine; Time Series Forecasting; Harmony Search
Código DOI: 10.21528/CBIC2013-303
Artigo em pdf: bricsccicbic2013_submission_303.pdf
Arquivo BibTex: bricsccicbic2013_submission_303.bib