Detection of absence epileptic seizures using support vector machine

Abstract

An application of support vector machine is presented as a tool for events detection in the electroencephalogram recorded from a patient clinically diagnosed with absence epilepsy. A comparison of five kernels is shown (linear, quadratic, polynomial, RBP and MLP) evaluating their efficiency for the detection of this epileptic event occurrence. The kernel with the best performance is the quadratic, with 99.43% accuracy in this specific case.