Machine Learning-based Alcoholism Detection Using Higuchi’s Fractal Dimension Features from EEG

Abstract

Alcoholism is a well-known mental health problem that affects behavior and emotional functioning. Excessive drinking of alcohol affects a person’s personal and social life in addition to causing mental diseases such as hallucinations, agitation, memory loss, difficulty in making decisions, and issues focusing. According to studies, employing various machine learning algorithms, alcoholism can be detected from different brain signals, particularly electroencephalography (EEG). This study investigates the classification performance of Neural Network (NN) with Higuchi’s fractal dimension (HFD) features from selected EEG channel data using the XGBoost algorithm. The classification accuracy is 94.79% using all channels while using selected channels it is also 94.16%. The selected channel-based NN classification model appears optimistic with a reduced number of channels without a significant reduction in accuracy.