BPSO Based Method for Screening of Alcoholism

Selection of optimal channels for alcoholic detection is a major issue in recent year. Advance knowledge of brain region or EEG channels, most affected by alcohol, will reduce the computational complexity and new EEG recording device can be designed using selected channels. In this paper, we propose complexity and nonlinearity features and ensemble subspace K NN classifier to differentiate alcoholics and nonalcohol's from visually evoked potential (VEP). Binary particle swarm optimization (BPSO) is used to select optimum number of channels that minimize classification errors. A novel fitness function is designed to use in optimization technique. Fitness function evaluated using classification error and selected channels. Experimental results show that optimal channel selected have biological significance associated with alcoholic person. Thus, the outcome of the proposed channel selectionmethodology can be used for the accurate and rapid classification of normal and alcoholic subjects.