Effectiveness of the Discrete to Continuous (DtC) Algorithm in Reducing EEG Dataset Dimensionality for Alcohol Use Disorder (AUD) Diagnosis

High dimensional datasets may be reduced through feature extraction and selection methods. A large dataset, like electroencephalograph measurements, may be difficult to interpret without specialized training and experience. Using a subset of relevant EEG channels to reduce EEG dataset dimension would improve classification models’ performance and computational efficiency. The Discrete to Continuous (DtC) approach was previously investigated to reduce EEG dataset dimensionality and improve analysis accuracy for alcoholism discrimination through selection of pertinent channels. AUD poses a major threat, and its effects can harm physical and mental health. Although AUD is a major issue and can be detected through EEG data, researchers have studied EEG signals less commonly to find pertinent channels for AUD detection and diagnosis. The present study sought to further evaluate the effectiveness of the DtC algorithm in selecting optimal EEG channels based on binary classification. We used time-domain features and logistic regression (LR). EEG recordings were retrieved from a public dataset. Experimental results using all EEG channels were compared to those employing only DtC-selected EEG channels, and to a third set including four EEG channels determined to be irrelevant for AUD discrimination according to the DtC method. Based on the results obtained, DtC-selected EEG set classification performances are higher than third set classification performances. Results clearly highlight significant achievements in all classification performance metrics for the DtC-selected EEG channels compared to the third set. Our findings encourage further investigations into the DtC approach to reduce EEG dataset dimension.