Decoding Three Different Preference Levels of Consumers Using Convolutional Neural Network: A Functional...

Area of Science:

• Neuromarketing
• Neuroimaging
• Machine Learning

Background:

• Consumer preference classification is crucial for evaluating marketing effectiveness.
• Functional near-infrared spectroscopy (fNIRS) measures brain activity, offering insights into consumer responses.
• Convolutional Neural Networks (CNNs) show promise in analyzing complex neuroimaging data.

Purpose of the Study:

• To decode consumer preference levels using a CNN with fNIRS data.
• To design a CNN-based fNIRS data analysis structure for high classification accuracy.
• To evaluate the performance of this method across different advertisement durations and consumer groups.

Main Methods:

• Eight healthy participants (4 female, 4 male) viewed 15, 30, and 60-second commercial videos.
• Cerebral hemodynamic responses were measured using fNIRS.
• CNN was employed to extract features (mean, peak, variance, kurtosis, skewness) for preference classification.

Main Results:

• The CNN-based fNIRS analysis achieved average classification accuracies of 84.3% (15s), 87.9% (30s), and 86.4% (60s) videos.
• The highest accuracy was 87.9% for 30-second videos.
• Male participants showed higher accuracy (88.4%) in distinguishing 'like' vs. 'dislike' preferences.

Conclusions:

• CNN-based fNIRS data analysis is an effective method for decoding consumer preferences.
• Advertisement duration impacts classification accuracy, with 30-second ads being optimal.
• The method demonstrates potential for objective neuromarketing insights into consumer intentions.