Electroencephalography (EEG) Signal Identification for Epilepsy Using Support Vector Machine (SVM)

Epilepsy causes seizures due to abnormal brain electrical activity. Electroencephalogram (EEG) signals are used in many medical applications, including epileptic seizure detection since they capture brain electrical activity. Neurologists have made inferences from visual observation. However, this technique takes time and depends on the reader’s skill. Thus, epilepsy detection must be automated, and machine learning can be used. This study concerns epilepsy EEG signal identification. Classifying preictal and ictal EEG signals and assessing the Support Vector Machine (SVM) method’s accuracy, precision, recall, and error value to identify epilepsy EEG signals. SVM will identify data based on ictal labels (1.0) and preictal (0.0) in the dataset with numerous parameters and data split ratio modifications. The first experiment evaluates the c parameter with a 7:3 data split ratio and the default rbf kernel. The training and validation stages produced accuracy, precision, and sensitivity values above 95% and an error rate below 0.5%. The second experiment tests kernel parameters with a 7:3 data split ratio. The C kernel parameter is set to its default value of 1. The training and validation stages produce over 77% accuracy, precision, and sensitivity with an error rate of 1.6%. The third experiment examines how changing the training-testing data split ratio affects the default kernel and parameter c. The training and validation steps produce 77% accuracy, precision, and recall with an error rate of 1.6%. Parameters and variations split data during testing to isolate the best data using parameters and ratios.