Automated Epilepsy Detection using Machine Learning Classifiers based on Entropy Features

Primary tool for studying human epileptic disease is an EEG recording. Visual Analysis of EEG recording is a very time consuming and tedious process. Therefore, a technique that can improve the precision of signal analysis is required. This paper suggests a method for improving the current state's diagnostic precision and offer a more accurate prediction of survivability. The proposed methodology uses an entropy based statistical approach to quantify epileptic EEG recordings which is more close to the behavior of epileptic EEG signals. Based on the extracted features, a model equipped with different classifier is suggested for improved inferences on the current state of the disease and projected survival rates in the future. Based on the performance metric, the Random Forest classifier outperforms in terms of accuracy (96%), specificity (96%) and sensitivity (95%).