DESCRIPTIVE VIEW OF FILTER APPROACH IN GA FOR STANDARD DATASETS OF IMAGES
PURUSHOTTAM DAS
Graphic Era Hill University, India
ANKUR SINGH BIST *
SVU, Gajraula, India
*Author to whom correspondence should be addressed.
Abstract
A genetic algorithm (GA), which uses mutual information (MI) for its fitness function, is presented for feature selection. Further, neural network (NN) is used for classification. Some standard i.e., Iris, Glass and Wine datasets are obtained from UCI repository. A domain dataset i.e., Rice dataset is also prepared from real world rice data. Finally, GA-MI-NN approach is presented and an existing Multi-SVM approach is also used. Multi-SVM is used for validating our results obtained on the taken datasets by GA-MI-NN approach.
Keywords: Feature selection, feature extraction, genetic algorithm, support vector machines, classification, neural network