Classification of Products Preference from EEG Signals using SVM Classifier

The investigation of the brain activities during the visualization of different commercial images can help better understand the brain activities and its application in neuromarketing. This work presents an evaluation of different EEG time and frequency domain features within different brain regions of interest under the support vector machine (SVM) classifier with the research's goal to determine the best features and brain regions corresponding to the customer feelings. An online available dataset of 25 users', using a 14 channel EEG system, responses to 42 products is used. The outputs included two classes: like and dislike. The data is preprocessed by filtration, independent component analysis (ICA), principle component analysis (PCA) and normalization. Sixteen features/feature groups are derived from the preprocessed data using a window size of one-second and a total of four seconds of EEG signal. The features are then studied in an SVM classifier. The accuracy of the classification varied between the different features ranging between 60.71% for the Alpha power and 66.25% for the signal's slope sign change (SSC) feature using all channels. Further, the frontal lobe of the brain gave higher accuracy in comparison with the other regions, and the left frontal lobe was more dominant than the right frontal lobe in relation to the product preference decision. The results suggest an improvement in the classification accuracy when applying ICA and PCA. The left frontal lobe has the potential to classify user decisions for future simplified systems.