Combining Facial Behavioral Cues, Eye Movements and EEG-Based Attention to Improve Prediction of Reading Failure

Abstract

Educational data mining can extract useful information from students reading data to predict their reading performance. This study aims to present a model for the identification of students who are likely to fail in reading. To improve predictive performance, multimodal data, facial behavioral cues, eye movements, and EEG-based attention, are combined to predict reading performance. Experimental results revealed that integrating students external subconscious behaviors and internal cognitive states was complementary to reading performance prediction. Also, the results demonstrated that adding internal data (EEG-based attention) presented a significant improvement in prediction effectiveness beyond using external data (facial behavioral cues and eye movements) alone. Hence, applying EEGbased attention could play an important role in raising the level of prediction quality. The findings contribute to the possibility of promoting students self-regulated learning and presenting appropriate feedback by instructors. From this, the exploring results support the potential for the improvement of learning engagement in reading.