Prediction of advertisement preference by fusing EEG response and sentiment analysis

Area of Science:

• Neuroscience
• Computer Science
• Marketing Science

Background:

• Accurate prediction of user engagement with video advertisements is crucial for marketing effectiveness.
• Existing methods often rely on subjective user feedback or limited sentiment analysis.
• Integrating objective physiological data with broader sentiment analysis offers a more comprehensive approach.

Purpose of the Study:

• To develop and evaluate a novel multimodal framework for predicting video advertisement ratings.
• To combine Electroencephalogram (EEG) signals with Natural Language Processing (NLP) sentiment analysis for enhanced prediction accuracy.
• To assess the improvement in prediction accuracy using the combined multimodal approach compared to EEG-only prediction.

Main Methods:

• Collected Electroencephalogram (EEG) data from participants watching video advertisements.
• Acquired and processed global textual comments using Natural Language Processing (NLP) for sentiment analysis.
• Developed a Random Forest regression model to predict ratings using EEG data.
• Fused EEG-based predictions with NLP-derived sentiment scores for a multimodal rating prediction.

Main Results:

• The multimodal approach demonstrated a lower Root Mean Square Error (RMSE) in rating prediction compared to using EEG data alone.
• The integration of physiological data (EEG) and public sentiment (NLP) significantly improved prediction accuracy.
• The proposed framework successfully combined user-specific physiological responses with global audience sentiment.

Conclusions:

• The multimodal framework integrating EEG and NLP offers a more accurate method for predicting video advertisement ratings.
• Combining objective user physiological responses with subjective public sentiment enhances predictive performance.
• This approach provides valuable insights for optimizing video advertisement content and strategy.