An Empirical Study on Comparison of Machine Learning Algorithms for Eye-State Classification Using EEG Data

Brain–computer interface (BCI) is an upspringing avenue that has dipped its hands in a variety of fields. BCI device collects brain waves from individuals in the form of electroencephalogram (EEG). The classification of eye-state such as closed or opened using electroencephalogram (EEG) data plays a crucial role in a variety of applications. This research endeavor is an empirical study on the popular machine learning algorithms—logistic regression, ElasticNet classifier, and support vector machine, with diverse kernels to arbitrate their efficiency in eye-state classification. The dataset comprised of continuous EEG measurements collected with Emotiv EEG Neuroheadset from individuals. The data was preprocessed to accommodate it to the classification algorithms. The findings revealed that support vector machine (SVM) with radial basis function (RBF) illustrated robustness in handling complex EEG data. Logistic regression promoted interpretability, while ElasticNet classifier offered a balanced approach. The accuracy of SVM with RBF kernel was 77%, while the accuracy of logistic regression and ElasticNet classifier was found to be 57.2% and 57.8%, respectively.