Robust Prediction of Alcoholism from EEG Signals Using Auto-Encoder

The diagnosis of alcoholism (substance abuse) is crucial due to its impact on individuals, society, and healthcare systems worldwide. Electroencephalography (EEG) signals have been shown to contain valuable information about the effects of alcoholism on the brain, and can potentially be used as a non-invasive tool for early detection and monitoring. In this study, we proposed an auto-encoder model for predicting alcoholism from EEG signals, which achieved robust and accurate performance even in the presence of noise and variability in the data. We evaluated the model using multiple classification algorithms, including Logistic Regression, Random Forest Classifier, AdaBoost Classifier, and AutoEncoders. The auto-encoder model (consisting of Bi-LSTMs and ReLU, Sigmoid activation for Dense and output layers respectively) outperformed the other algorithms, achieving an accuracy of 65.5% on the test set, while Logistic Regression achieved an accuracy of 55%, Random Forest Classifier achieved 58%, and AdaBoost Classifier achieved 58.4%. These results demonstrate the effectiveness of auto-encoder models for biomedical signal processing and provide a promising avenue for future research in this area, including the investigation of more complex architectures and feature engineering techniques, and the use of larger and more diverse datasets for training and evaluation. This research contributes to the fields of biomedical signal processing, neurophysiology, and neuroimaging, offering insights into brain activity related to alcoholism and highlighting the importance of non-invasive diagnosis and early detection in tackling alcoholism disorders and their societal impact.