Abstract
This study estimates cognitive effort based on functional near-infraredspectroscopy data and performance scores using a hybrid DeepNet model. Theestimation of cognitive effort enables educators to modify material to enhancelearning effectiveness and student engagement. In this study, we collectedoxygenated hemoglobin using functional near-infrared spectroscopy during aneducational quiz game. Participants (n=16) responded to 16 questions in aUnity-based educational game, each within a 30-second response time limit. Weused DeepNet models to predict the performance score from the oxygenatedhemoglobin, and compared traditional machine learning and DeepNet models todetermine which approach provides better accuracy in predicting performancescores. The result shows that the proposed CNN-GRU gives better performancewith 73% than other models. After the prediction, we used the predicted scoreand the oxygenated hemoglobin to observe cognitive effort by calculatingrelative neural efficiency and involvement in our test cases. Our result showsthat even with moderate accuracy, the predicted cognitive effort closely followthe actual trends. This findings can be helpful in designing and improvinglearning environments and provide valuable insights into learning materials.