Epilepsy Classification Using Support Vector Machine with Frequency Domain Feature Extraction

Abstract

Epilepsy is a brain-related condition characterized by abnormal electrical activity in the brain. This condition can be identified by observing EEG signals, which record the brain's electrical activity. Automatically detecting seizures using EEG signals helps doctors diagnose the condition more quickly and accurately. In this study, a method is proposed that uses a Support Vector Machine (SVM) to classify EEG signals. The features used for classification are extracted from the frequency domain using a technique called Fast Fourier Transform (FFT). The dataset used is called the UCI Epileptic Seizure Recognition Dataset, which includes 11,500 EEG samples divided into five classes. These samples are then categorized into two main types: seizures and non-seizures. The research process includes data preprocessing with MinMaxScaler normalization, feature extraction using FFT, and data classification using SVM with varying numbers of features. Model performance is measured using several metrics, including accuracy, precision, recall, F1 score, and ROC-AUC. The results showed that the use of 21 features with the SVM model provided the best performance, with an accuracy of 97.7%, a precision of 93.2%, a recall of 95.4%, an F1 score of 94.3%, and an AUC value of 0.9930. These results are better than previous studies using similar methods, indicating that the combination of FFT and SVM is effective for detecting epilepsy using EEG signals. These findings help build a more reliable system for medical diagnosis and highlight the importance of using balanced evaluation measures in healthcare.