Logistic Regression Classifier with Hybrid Dimensionality Reduction Techniques for Epilepsy Detection from EEG Signals

Epilepsy is a fatal disease of the nervous system that affects a big part of the world's population. Electroencephalography (EEG) is the tool that is most often used in clinical settings to measure how the brain works. EEG is often used to diagnose and treat a wide range of health problems, such as mental exhaustion, epileptic seizures, coma, schizophrenia, and sleep problems. The large dimensionality of data is one of the problems with epilepsy detection that is discussed in this work. When attempting to analyze the data, a high dimensionality might be problematic. The primary purpose of lowering the dimensionality of the EEG data is to augment the classification results. In this paper, the dimensionality of the EEG data can be reduced by the use of procedures such as dimensionality reduction and feature selection. For dimensionality reduction, the autoencoder and Hessian LLE method are employed. As a technique of feature selection, the PAC Bayesian algorithm is employed. Logistic Regression classifier is used to detect epilepsy. Results show that the when Hessian LLE with PAC-Bayesian is identified with logistic regression gives the best accuracy of 96.87%.