Improved EEG-Based Epileptic Seizure Classification Using Fusion of Machine Learning Model with Statistical Feature Integration

Epilepsy is a neurological disorder characterized by recurrent seizures, often diagnosed through the analysis of electroencephalogram (EEG) signals. In this study, we propose a machine learning-based approach to classify EEG data into different brain states, focusing on epileptic seizure detection. Discrete Wavelet Transform (DWT) is employed to extract relevant time-frequency domain features from the EEG signals. Additionally, a set of statistical features—mean, standard deviation, variance, energy, entropy, coefficient of variation, root mean square, skewness, and kurtosis are incorporated to further enhance classification performance. Various machine learning models such as Random Forest, KNN, and LightGBM were evaluated. Results show that combining DWT with statistical features significantly improves the accuracy and robustness of the classification compared to using DWT alone. For example, the Random Forest model achieved an accuracy of 84.85% with DWT alone, which increased to 84.88% when combined with statistical features. Similarly, LightGBM improved from 84.87% to 85.10% in accuracy. This approach could serve as a valuable tool for aiding in the diagnosis and treatment of epilepsy.