Selecting the relevant features from the electroencephalogram (EEG) data that can differentiate normal and epileptic classes of data with promising accuracy is a multifaceted problem. Feature selection accounts for recognize a subset of features and in consequence eliminate the irrelevant features. In this paper, we propose an optimization approach that performs the feature selection by considering the "chaotic" version of firefly optimizer, which is a swarm intelligence family of algorithms

more »

... t mimics the nature inspired flashing lights mechanism of fireflies. The balance between exploration of the search space and exploitation of the best solutions is a challenge in multi-objective optimization, to maximize the eminence of the data-training fitting model with reduced feature set. In this paper, chaotic map is used to produce the chaotic sequence and used to control the feature optimization process. The purpose of chaotic maps is to determine the light absorption coefficient of the firefly algorithm (FFA). We propose Joint Logistic-tent map (JLTM) based improved chaotic firefly algorithm (ICFFA) to implement the feature selection followed by Multi-Class Support Vector Machine (MSVM) for evaluating the classification accuracy. We generate the chaos streams using various chaotic maps. The results have shown that the JLTM is recognized as being the most important chaotic map to increase the overall quality of the ICFFA performance. The experimental results prove the JLTM based ICFFA leads to improved classification accuracies when compared with state-of-the-art methods.