Detection of Epilepsy in EEG Signal using Dimension Reduction based Stacking Classifier

Epilepsy is a neurological illness that occurs in the brain and generates seizures. The life of the patient can be saved if diagnose at an early stage. Based on the reports of WHO, near about 50 million people suffer by this disease. EEG is a monitoring tool used to extract the brain signals and different machine learning techniques help in analyzing these signals for the prediction of epilepsy. Several traditional algorithms prescribed in the earlier papers for this purpose lack by the issue of low accuracy, precision values, high complexity in algorithm design, time constraint. Therefore an ensemble "Stacking Classifier" technique is proposed in this study to enhance the overall performance and minimize the chance of misclassification of an instance. The classifiers used in this study include Logistic Regression(LR), K-Nearest Neighbors(KNN), Ridge classifier (RC), Decision Tree(DT), Gaussian Naïve Bayes(NB), Support vector machine (SVM) and Stochastic Gradient Descent (SGDC).The predictions of these models are aggregated and applied on Light Gradient Boosting Machine (LGBM) model which act as Meta classifier to give the final prediction. Furthermore Principal Component Analysis (PCA) is employed for reducing the dimension of the dataset. When it comes to Accuracy (Acc), Precision(P), Recall(R) , and F1 score(F1), the proposed Stacking classifier achieves 96.69%, 90.55%, 93.41% and 91.96% respectively. Additional matrices like Critical Success Index (CSI) and Cohen’s Kappa (K) also computed to give additional information about the model and the proposed one achieves 85.11% and 89.88% respectively.