EEG-Based Tool for Prediction of University Students’ Cognitive Performance in the Classroom

Ramírez-Moreno, Mauricio A.; Díaz-Padilla, Mariana; Valenzuela-Gómez, Karla D.; Vargas-Martínez, Adriana; Tudón-Martínez, Juan C.; Morales-Menendez, Rubén; Ramírez-Mendoza, Ricardo A.; Pérez-Henríquez, Blas L.; Lozoya-Santos, Jorge de J.

This study presents a neuroengineering-based machine learning tool developed to predict students’ performance under different learning modalities. Neuroengineering tools are used to predict the learning performance obtained through two different modalities: text and video. Electroencephalographic signals were recorded in the two groups during learning tasks, and performance was evaluated with tests. The results show the video group obtained a better performance than the text group. A correlation analysis was implemented to find the most relevant features to predict students’ performance, and to design the machine learning tool. This analysis showed a negative correlation between students’ performance and the (theta/alpha) ratio, and delta power, which are indicative of mental fatigue and drowsiness, respectively. These results indicate that users in a non-fatigued and well-rested state performed better during learning tasks. The designed tool obtained 85% precision at predicting learning performance, as well as correctly identifying the video group as the most efficient modality.