Abstract:

Students with special educational needs (SEN) often face difficulties that include hyperactivity, short attention span, and emotional instability, which can impact their academic, social, and personal growth. Applied Behavior Analysis (ABA) is an evidence-based intervention that uses the principles of reinforcement and stimulus control to promote desirable behavioral outcomes. While ABA interventions are systematic and based on empirical evidence, very little research has been conducted on predicting behavior change through advanced data-driven methodologies. This study seeks to improve the ABA practice through a multimodal learning analytics, MMLA, as well as machine learning (ML) technique for the purpose of predicting behavior change among SEN students. The work developed a multimodal data collection system that gathers data from 1,130 sessions of ABA therapy comprising ambient, physiological, as well as motion data, and statistical analysis shows sensor and wearable data significantly improve accuracy in comparison to traditional educational data. Furthermore, ML models including DNN effectively predict changes in behavior with performance benchmarking of effectiveness against related works. This research brings new insights to ABA practices in integrating IoT technologies and advanced analytics into behavioral interventions. The results have practical implications for the improvement of student outcomes in SEN and open a ground for future work on MMLA's application in special education.

Description:

.