[Paper Review] Understanding Consumer Preferences for Movie Trailers from EEG using Machine Learning

Pankaj Pandey, Raunak Swarnkar+2 more|arXiv (Cornell University)|Jul 21, 2020

EEG and Brain-Computer Interfaces16 references4 citations

TL;DR

This study uses EEG signals and machine learning to predict consumer preferences for movie trailers with 72% accuracy, demonstrating that neural correlates—particularly in beta frequency bands—can effectively decode individual viewing preferences beyond self-reported ratings.

ABSTRACT

Neuromarketing aims to understand consumer behavior using neuroscience. Brain imaging tools such as EEG have been used to better understand consumer behavior that goes beyond self-report measures which can be a more accurate measure to understand how and why consumers prefer choosing one product over another. Previous studies have shown that consumer preferences can be effectively predicted by understanding changes in evoked responses as captured by EEG. However, understanding ordered preference of choices was not studied earlier. In this study, we try to decipher the evoked responses using EEG while participants were presented with naturalistic stimuli i.e. movie trailers. Using Machine Learning tech niques to mine the patterns in EEG signals, we predicted the movie rating with more than above-chance, 72% accuracy. Our research shows that neural correlates can be an effective predictor of consumer choices and can significantly enhance our understanding of consumer behavior.

Motivation & Objective

• To investigate whether EEG signals can predict ordered consumer preferences for movie trailers.
• To explore the neural correlates of consumer choice using non-invasive brain imaging and machine learning.
• To overcome limitations of self-report measures in neuromarketing by leveraging objective neural data.
• To evaluate the effectiveness of machine learning in decoding brain responses to naturalistic stimuli like movie trailers.

Proposed method

• Collected 128-channel EEG data from 18 participants viewing 12 movie trailers using a Net Station device.
• Preprocessed EEG signals using EEGLab and artifact removal techniques (EOG, EMG, movement, contact issues).
• Applied Discrete Wavelet Transform (DWT) with db-8 wavelet to extract Power and Entropy features across five frequency bands (0–60 Hz).
• Used Recursive Feature Elimination (RFE) and Sequential Backward Selection (SBS) for feature reduction.
• Trained nine machine learning classifiers (e.g., kNN, Random Forest, SVM) on the reduced feature set with 10-fold cross-validation.
• Optimized hyperparameters iteratively using Python and evaluated performance on a test set of 65 samples.

Experimental results

Research questions

• RQ1 Can EEG signals predict individual movie trailer preferences with higher accuracy than self-reported ratings?
• RQ2 Which EEG frequency bands and features (Power, Entropy) are most predictive of consumer preference?
• RQ3 How effective are different machine learning classifiers in decoding neural patterns associated with preference?
• RQ4 To what extent do neural correlates outperform behavioral self-reports in predicting ordered preference rankings?

Key findings

• The k-Nearest Neighbors (kNN) classifier with Recursive Feature Elimination achieved the highest test accuracy of 72.37% in predicting movie trailer ratings.
• All five frequency bands (delta, theta, alpha, beta, gamma) contributed discriminative information for classification.
• Beta band activity showed strong predictive power for like/dislike preferences, consistent with prior studies on vmPFC and reward processing.
• Feature elimination using SBS and RFE significantly improved model performance by reducing dimensionality and noise.
• The small sample size (n=216) relative to high-dimensional features (n-large-p problem) posed a major challenge, limiting model generalization.
• While PCA did not improve performance, non-linear dimensionality reduction (e.g., t-SNE, UMAP) and deep learning techniques were identified as promising future directions.