Classification Method of EEG Based on Evolutionary Algorithm and Random Forest for Detection of Epilepsy

Abstract

Epilepsy is a difficult problem that has puzzled the medical profession for a long time. The complexity, randomness, non-stationarity and nonlinearity of EEG signal of epilepsy bring great challenge to the detection of epilepsy. The study of epilepsy is an important subject of neutral system diseases. For automatic epilepsy detection system, the accuracy of identifying epilepsy and predicting epilepsy is of great significance to the treatment of doctors and the recovery of patients. This paper proposes the mixed feature extraction to extract the feature by mixture of time-domain method and nonlinear analysis method, and the extracted feature is optimized using evolutionary optimization algorithm, and finally train the epilepsy classifier by utilizing the optimized features through the Random forest algorithm. In the experiment, the accuracies of two-classification problems and three-classification problems respectively reach 99.2% and 98.1%. The results of cross-over experiment for many times show that, the method is of effectiveness in the classified feature extraction aiming at epilepsy brain wave.