Prediction Accuracy of Eye-Open State using WEKA Algorithms

Abstract

Purpose: Brain diseases are reflected in the pattern of brain-waves recorded using electroencephalography (EEG). We aimed to evaluate the prediction accuracy of machine learning algorithms embedded in WEKA software tool applied to the EEG eye-state signal dataset.  Methods: The eye-state dataset was retrieved from UCI ML repository, and it consists of 14980 samples (instances), 15 attributes (electrodes), and each instance was one continuous EEG measurement made within 117 seconds. The two classes in the dataset are '1', indicating the eye-closed state and '0' the eye-open state. The prediction accuracy of eye-closed and eye-open was done with machine learning algorithms incorporated in WEKA software tool. Results: The best statistical performance evaluation measure was observed in this study for the classifiers viz., Random Forest, Random Tree, J48, Bagging and Decision table. Random Forest predicted the edited test dataset in the ratio of 7:3 (correct : incorrect). Conclusion: Among the five classifiers, Random Forest and Bagging gave significant performance (‘v’) while analyzed in the ‘experimenter’ environment in WEKA.