EEG-Based Detection of Alcoholic Status with Heuristic Classification Techniques

This paper discusses a comparative study of three heuristic algorithm-based classifiers namely, Firefly Algorithm, Harmonic Search, and Particle Swarm Optimization (PSO) for classifying the alcoholic status from EEG signals. These classifiers were evaluated on EEG samples processed through two dimensionality reduction (DR) techniques: Ridge Regression and Fuzzy C-Means and Clustering. Assessment of the efficacy of classifiers was evaluated by using sensitivity, specificity, accuracy, F1 Score, Measure of Correct Classification (MCC) and error rate. The classification results show that the PSO classifier integrated with Fuzzy C Means Cluster resulted in the highest accuracy of 94.26 with a lower error rate of 5.74 hence highlights good classification ability for this hybrid. The performance of the Firefly and Harmonic Search classifiers were moderate with worse case accuracy level of 77% with the firefly reaching an accuracy of 77% when used with the ridge regression but the firefly algorithm was seen to be inconsistent with the DR techniques used. In general, the Fuzzy C-Means Clustering was identified as the best DR technique that improved the performance of classifiers due to feature extraction and improvement of clustering.