Comprehensive Evaluation Of Machine Learning Algorithms For Epileptic Seizure Categorizationn With PCA

Abstract

One of the most dreaded disorders of the central nervous system, epilepsy is linked to abnormal activity that explodes in the brain. Neurologists employ a well-known method called electroencephalography (EEG), which records the electrical anomalies emerging from the brain, to diagnose this illness. Due to the large volume, complexity, and nondeterministic nature of the obtained signal data, interpreting these recordings requires an expert, who is hard to come by in developing nations. Therefore, in order to improve computer aided diagnosis (CAD) solutions, a comprehensive comparison of four crucial machine learning (ML) algorithms—Logistic Regression (LR), K-Nearest Neighbour (K-NN), Decision Tree (DT), and Naive Bayes (NB)—for seizure categorization is offered. Using the Time frequency domain (TFD) approach, which is a member of Cohen's distribution class, pertinent properties are chosen from the EEG dataset. Model efficacy is analysed both pre- and post-introduction of principal component analysis (PCA) to the dataset. Our results indicate that as the complexity of the dataset is reduced, the classification accuracy enhances. The NB classifier performs better than the other classifiers, which makes it the most appropriate for classifying epileptic instances.