Ordered Weighted Aggregation Operators based Statistical features for Seizure Classification using EEG

Epilepsy is a chronic neurological disorder that affects a significant proportion of the global population. A proper identification is essential for successfully diagnosis, treatment, and controlling epilepsy. Clinically non-invasive electroencephalography (EEG) is primarily employed for observing these conditions, allowing the diagnosis of the disorder. The visual assessment of EEG signals is tedious and time-intensive. In this regard, several automated techniques are proposed; however, it should be noted that the computational complexity associated with training the model is higher for these methods. Additionally, the extracted features from these methods contribute equally, which can result in neglecting higher deviation levels in the signal to identify the disorder. These, in turn, may lead to false positive results. To overcome this and enhance the performance of the decision, we propose a novel ordered weighted aggregation (OWA) based statistical feature, which will serve as the feature set for the classification model to identify the disorder. The OWA based statistical features prioritize the necessary components by providing the weight during the evaluation of the statistical attributes and are used as input for the machine learning models. We tested the model’s performance over the University of Bonn dataset to show the model’s efficacy. The proposed approach’s performance accuracy is higher compared to the state-of-the-art methods. Finally, the study highlights the significance of OWA-based statistical features in the framework by using SHAP values to showcase the impact of these features.