<p>Early detection of alcoholism allows for early intervention and treatment that increases the chances of successful treatment and recovery. The longer the alcoholism goes undetected, the greater the risk of developing numerous health risks, including liver disease, cardiovascular problems, gastrointestinal issues, neurological damage, and mental health disorders. Detecting alcoholism enables healthcare professionals to address these risks promptly and provide appropriate medical interventions and support. This article proposes a novel method combining the empirical wavelet transform and the Fourier decomposition method for automated detection of alcoholism using electroencephalogram (EEG) signals. The new method, called the empirical Fourier decomposition (EFD) method, which combines the improved Fourier segmentation technique and zero phase filter bank concepts. The EFD method is applied to decompose the EEG signals into sub-band wave components. Features such as mean, standard deviation, kurtosis, line length, Hjorth parameters, log energy entropy, and norm energy entropy are extracted from each component. To select the optimal features, the horse-herd optimization algorithm (HHOA) is employed to identify the best feature subset. The experiments have been carried out using classifiers, such as least squares support vector machine (LS-SVM), random forest (RF), and k-nearest neighbor (k-NN). The proposed approach provides average accuracies of 99.01% with RF, 98.36% with LS-SVM, and 98.14% with the k-NN classifiers. The experimental results show that the proposed method outperforms the existing methods and can be adopted for real-time alcoholism detection. The decomposition algorithm presented demonstrates enhanced performance across both standard and acquired datasets, indicating its robustness and generalizability in alcoholism detection.</p>