Confusion-Emotion-Recognition from EEG Signals from Artificial Intelligence

Recognizing learners' confusion is essential for educators as it greatly impacts learning. While confusion can initially motivate students, prolonged confusion can hinder the learning process.

Educators typically identify confusion based on their own understanding, but advances in Affective Computing have led to the development of computerized systems that use psychophysiological signals (e.g., EEG) and Machine Learning (ML) models to recognize learner emotions.

However, current research lacks sufficient focus on designing confusion emotion recognition systems suitable for real-world learning environments In this research, two major identified research gaps were identified and addressed. They were:
a) Reduced model performance in existing confusion emotion models due to Bias – Variance Dilemma

b)Limited R&D on cost-sensitive model which may lead to cost inefficiency in existing confusion emotion models

Addressing the research gaps aided in developing confusion emotion model that is both performance and cost efficient.

The performance was increased by hyperparameter tuning the utilized supervised ML algorithms using Genetic Algorithm and was made cost efficient by training a heterogenous ensemble model that had reduced misclassification costs.