Research on Alcoholism EEG Signal Classification Based on FE-IIPSO-SVM

Alcohol abuse poses significant health risks and social hazards. In recent years, the identification of alcohol-dependent individuals through EEG signal analysis has emerged as a prominent research focus. Traditional time-frequency EEG classification methods primarily concentrate on local signal characteristics, failing to adequately explore inter-regional connectivity features across different brain areas. This limitation results in an incomplete representation of brain activity, leading to insufficient understanding of neural mechanisms and suboptimal classification accuracy. This study investigates the differences between alcohol-dependent patients and healthy controls based on brain network characteristics. Given the high number of EEG channels and the excessive dimensionality of cross-frequency fusion features, manual classification experiments with different feature combinations are practically infeasible. To address this challenge, we propose an Feature-Enhanced Improved Immune Particle Swarm Optimization-Support Vector Machine algorithm for brain network fusion feature classification. The method introduces a novel particle mutation strategy and constructs an immune memory bank alongside a parent population to preserve high-fitness particles. By encoding feature combinations and employing the enhanced immune particle swarm algorithm to traverse and screen potential combinations, we identify the optimal feature set with the highest classification accuracy. Experimental results demonstrate that the proposed method achieves a classification accuracy of 92.2%, significantly outperforming traditional SVM algorithms in both accuracy and computational efficiency. This approach not only enhances classification performance but also substantially reduces processing time, offering a robust framework for EEG-based alcohol dependence identification.