EEG Brainwave Emotion Detection Using Stacked Ensembling Method

Emotion classification based on brain signals is popular in the Brain-machine interface. In BMI, machine learning techniques have proved to show better performance than traditional classification methods. The study examines a dataset collected using various signals that are recorded as a classification of BMI systems. The study implements stacking, an ensembling technique for emotion detection to achieve improved accuracy. Eight base models namely, Support Vector Classifier, Xgboost, Lightgbm, Random Forest, Deep neural networks, Logistic Regression, and K-Nearest neighbors are trained. Accuracies ranging from 91.36% to 96.06% were achieved using the base models. The base models and the dataset form the input of the proposed stacked meta-model. Accuracy as high as 97.0% was obtained using the proposed stacked meta-model.