Predicting Consumer Choices Using Brain Signals and a Hybrid Gray Wolf–Cheetah Algorithm

Abstract

The diversity of customer decision-making and product preferences combine to form a marketer's nightmare. In this work, we introduce a new neuromarketing method for predicting consumer choices based on brain signals. Participants were 25 healthy volunteers (1838 years old) who observed 14 items while their electroencephalogram (EEG) was recorded. In this paper, a hybrid feature selection method is proposed on the basis of the parallel population strategy of the Gray Wolf Optimizer and the Cheetah Optimization Algorithm. In contrast to the existing approaches, with this hybrid model the exploration ability of the GWO is combined with the exploitation rate of Cheetah Algorithm, where the convergence rate and the feature relevance on both of them are improved simultaneously. Spectra high order, a method capturing non-linear and complex pattern in EEG signals, was applied for the initial feature extraction, resulting 742 features. The proposed algorithm further yielded a reduced set of 174 key features. It was found that the average prediction accuracy of the model in predicting users’ product choices was 76.84%, 4.92% higher than that of baseline methods. These findings show the promise of this method in improving targeted advertising and creating a more personalized consumer experience.