A new evolutionary neural network classifier

Thammano, Arit; Meengen, Asavin

doi:10.1007/11430919_31

A new evolutionary neural network classifier

Author(s)

Thammano, Arit

Meengen, Asavin

Date Issued

January 1, 2005

Type

Conference Paper

DOI

10.1007/11430919_31

Abstract

This paper proposes two new concepts: (1) the new evolutionary algorithm and (2) the new approach to deal with the classification problems by applying the concepts of the fuzzy c-means algorithm and the evolutionary algorithm to the artificial neural network. During training, the fuzzy c-means algorithm is initially used to form the clusters in the cluster layer; then the evolutionary algorithm is employed to optimize those clusters and their parameters. During testing, the class whose cluster node returns the maximum output value is the result of the prediction. This proposed model has been benchmarked against the standard backpropagation neural network, the fuzzy ARTMAP, C4.5, and CART. The results on six benchmark problems are very encouraging. © Springer-Verlag Berlin Heidelberg 2005.

Citation

Lecture Notes in Computer Science Including Subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics, 3518 LNAI, 249-255, 2005