Multimodal Learning Analytics for Predicting Behavioural Change in Special Education Students

Abstract: To predict behavioural changes in children with special educational needs (SEN), this study introduces a novel multimodal learning analytics approach. Our algorithm provides great accuracy in detecting possible behavioural shifts by combining behavioural, emotional, and physiological signs with conventional academic criteria. More than 3,000 SEN students provided us with extensive data, such as attendance logs, emotional states, social interaction measures, attention span assessments, and engagement statistics on digital platforms. Our model was able to predict behavioural changes with 87% accuracy by using a Random Forest Classifier that was optimised by grid search cross-validation. Teachers can use these predictions for prompt interventions by putting in place a web-based program. Our results show that the most important determinants of behavioural change in SEN students are emotional state, attention span, and social interaction skills. This study offers helpful resources for assisting students with a variety of learning needs and contributes to the growing body of information in precision education. Keywords: Special education, Multimodal learning analytics, Behavioural prediction, educational technology, Student support systems, attention deficit hyperactivity disorder