This study aims to automatically classify the eye openness state (open/closed) of individuals from electroencephalography (EEG) signals. In the classification process, based on the knowledge that EEG signals reflect short-term cognitive states, the EEG Eye State dataset is used. The dataset contains 14,980 samples from 14 EEG channels and the eye state is labelled according to the binary classification problem. Within the scope of the preprocessing steps for the data, the scaling process was performed and then the classification model was created. In the modelling process, the Extra Trees Classifier (ETC) algorithm, which is an ensemble learning method based on decision trees, was preferred. The performance of the model was evaluated by 10-fold cross-validation method; accuracy, precision, sensitivity and F1 score metrics were calculated at each layer. The findings revealed that the model performed well in all metrics. In particular, the highest F1 score was achieved in Fold 1, and the width of the area under the ROC curve (AUC) confirmed the discriminative power of the model. In addition, in the feature importance analysis, it was observed that the signals obtained from occipital and parietal regions contributed more to the classification process. The results show that traditional machine learning algorithms, together with appropriate preprocessing strategies, can produce effective classification outputs on EEG data. This study contributes to the academic literature on EEG-based eye state detection and provides a meaningful basis for applications such as human-computer interaction, attention monitoring systems and neurocognitive assessment.
This study aims to automatically classify the eye openness state (open/closed) of individuals from electroencephalography (EEG) signals. In the classification process, based on the knowledge that EEG signals reflect short-term cognitive states, the EEG Eye State dataset is used. The dataset contains 14,980 samples from 14 EEG channels and the eye state is labelled according to the binary classification problem. Within the scope of the preprocessing steps for the data, the scaling process was performed and then the classification model was created. In the modelling process, the Extra Trees Classifier (ETC) algorithm, which is an ensemble learning method based on decision trees, was preferred. The performance of the model was evaluated by 10-fold cross-validation method; accuracy, precision, sensitivity and F1 score metrics were calculated at each layer. The findings revealed that the model performed well in all metrics. In particular, the highest F1 score was achieved in Fold 1, and the width of the area under the ROC curve (AUC) confirmed the discriminative power of the model. In addition, in the feature importance analysis, it was observed that the signals obtained from occipital and parietal regions contributed more to the classification process. The results show that traditional machine learning algorithms, together with appropriate preprocessing strategies, can produce effective classification outputs on EEG data. This study contributes to the academic literature on EEG-based eye state detection and provides a meaningful basis for applications such as human-computer interaction, attention monitoring systems and neurocognitive assessment.
Primary Language | Turkish |
---|---|
Subjects | Electrical Engineering (Other) |
Journal Section | Research Article |
Authors | |
Early Pub Date | July 1, 2025 |
Publication Date | |
Submission Date | May 22, 2025 |
Acceptance Date | June 12, 2025 |
Published in Issue | Year 2025 Volume: 15 Issue: 1 |
All articles published by EJT are licensed under the Creative Commons Attribution 4.0 International License. This permits anyone to copy, redistribute, remix, transmit and adapt the work provided the original work and source is appropriately cited.