EEG Signal Classification for Management of Epilepsy using RBF Kernel-Based Support Vector Machine

The human brain is a remarkable and complex organ that controls all the human body parts. The electroencephalography (EEG) is a cheaper and more readily available technique. By leveraging from the abilities of machine learning (ML), we can enhance EEG analysis to uncover hidden patterns, and advance neuroscience research. It can also play key roles in diagnostics, and personalized healthcare. EEG signals are challenging to analyze due to their nonlinear, non-stationary nature and susceptibility to noise. In order to address these issues, wavelet technique is employed for denoising EEG signals, offering advantages such as multi-resolution analysis, time-frequency localization, noise reduction, and signal feature preservation. In this paper, we have proposed radial basis function (RBF) with kernel based support vector machine (RSVM) to classify EEG signal for epilepsy and non-epilepsy cases. The proposed method combines wavelet decomposition with SVM. Using an RBF kernel for classification, we achieve higher accuracy, sensitivity, and specificity compared to other ML models. The findings demonstrate that RSVM exhibits the best level of classification accuracy (99.33%), specificity (100%), and sensitivity (99<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">%</sup>) when distinguishing between normal and epileptic patients.