Optimal feature selection using PSO with SVM for epileptic EEG classification

Abstract

Electroencephalogram (EEG) signals usually have a high dimensionality which makes it difficult for classifiers to learn the difference of various classes in the underlying pattern in the signal. Feature selection is one of the most important factors which can influence the classification accuracy rate. The objective of this research is to simultaneously optimise the parameters and feature subset. A novel scheme to detect epileptic seizures from electroencephalogram (EEG) is proposed in this paper. This scheme is based on discrete wavelet packet transform and uses the transform coefficients to compute energy, entropy, kurtosis, skewness, mean, median and standard deviation to form feature vector for classification. Optimal features are selected and parameters are optimised using Particle Swarm Optimisation (PSO) with support vector machine as a classifier for creating objective function values for the PSO. Clinical EEG data from epileptic and normal subjects are used in the experiment. To evaluate the efficacy of the proposed scheme, a tenfold cross-validation is implemented, and the detection rate is found 100% accurate with 100% of sensitivity and specificity for the data under consideration.