Kullback-Leibler divergence for Efficient EEG Feature Selections

This paper presents a new feature selection technique mainly based on the Kullback-Leibler (KL) divergence to diagnose epileptic seizures in the EEG dataset. Utilizing high-dimensional medical datasets necessitates implementing feature selection as a crucial procedure for the early detection of diseases to safeguard public health. Furthermore, high-dimensional data significantly impacts the predictive accuracy of machine learning algorithms and amplifies system complexity, thereby diminishing the efficiency of outcomes. Furthermore, after feature selection, some features do not play any significant role in effective classification. Thus, EEG classification followed by the selection of optimal features can produce better results. From a mathematical perspective, the proposed model undertakes the computation of signal similarities, selects optimal features characterized by high similarity, and eliminates redundant features. The evaluation of the system's performance was conducted using the EEG Bonn University dataset. Furthermore, metrics such as accuracy, precision, and recall were employed for assessment, with comparative analysis against existing methodologies. The experimental findings substantiated the efficacy of the proposed model as a valuable instrument for optimal feature selection in EEG data analysis. Notably, the suggested model achieved a 100% accuracy.