Alcoholism Identification From Variational Mode Decomposed EEG Signal Using Machine Learning Techniques

Alcohol addiction, which results in physical and psychological reliance, is defined as an inability to control drinking. Abuse of alcohol can increase the risk of getting certain malignancies. Alcohol use disorders tend to Cirrhosis and can harm the brain. Analyzing Electroencephalogram (EEG) signals is one of the most popular approaches for spotting alcoholism. This paper presents an alcoholic classification approach using variational mode decomposition (VMD) method and a Light Gradient Boosting machine (LightGBM) classifier. The raw EEG signals are preprocessed and decomposed into Variational Mode Functions (VMFs) using the VMD method. The correlation-based approach is used to select the most significant VMD modes and the spectral moment and nonlinear features are determined. The features collected are employed in k-fold cross-validation technique, and the alcoholic and control signals are distinguished by applying the XGBoost and LightGBM classifiers. The proposed model is evaluated with the publicly available UCI Alcoholic EEG dataset and evaluates the precision, accuracy, recall, and F1-score measures. The proposed approach performs significantly with the LightGBM classifier and reports precision, accuracy, recall, and F1-score measures are 99.88%,98.81%, 98.23%, and 99.04 respectively.