3.3.6. Naive Bayes

The Naive Bayes algorithm is a model based on Bayes’ theorem to predict outcomes by calculating the likelihood, or probability, of each possible result based on the data. This model assumes independence between features, in which it doesn’t have to account for interactions between features, which simplifies calculations and makes Naive Bayes efficient and scalable. Due to its simplicity, Naive Bayes is often effective with smaller datasets. However, it can face challenges when features are correlated, which is often the case in emotion recognition, where we want to leverage the relationship between EEG and ECG signals.

Dua also employed a Gaussian Naive Bayes classifier to recognize emotions from combined ECG and facial data, achieving an overall accuracy of 65.5%. Cao et al. proposed the LUR (Learn, Unlearn, and Relearn) model for EEG-based emotion recognition, which leverages Naive Bayes as one option for classifying emotions, achieving 66% accuracy for valence and 71% accuracy for arousal on DREAMER dataset.