A Binary classifier based on Firefly Algorithm

Abstract

This work implements the Firefly algorithm (FA) to find the best decision hyper-plane in the feature space. The proposed classifier uses a cross-validation of a 10-fold portioning for the training and the testing phases used for classification. Five pattern recognition binary benchmark problems with different feature vector dimensions are used to demonstrate the effectiveness of the proposed classifier. We compare the FA classifier results with those of other approaches through two experiments. The experimental results indicated that FA classifier is a competitive classification technique. The FA shows better results in three out of the four tested datasets used in the second experiment