PREDICTING BEHAVIOR CHANGE IN STUDENTS WITH SPECIAL EDUCATION NEEDS USING MULTIMODAL LEARNING ANALYTICS

Abstract

Students with special education needs 
(SEN), whose behavior and learning are 
particularly 
sensitive 
to 
their 
bodily 
conditions and surrounding surroundings, 
have new options due to the availability of 
educational data in innovative methods and 
forms. In order to elucidate the underlying 
educational insights, multimodal learning 
analytics (MMLA) collects student and 
learning environment data in a variety of 
modalities and analyzes it. In this study, we 
used MMLA to predict how SEN kids 
would behave after they get applied 
behavior analysis (ABA) treatment. ABA 
therapy is a kind of special education 
intervention that tries to cure behavioral 
issues and promote positive behavior 
changes. Here, we demonstrate how our 
deep neural network and machine learning 
models can optimally predict the behavior 
change of SEN kids with 98% accuracy and 
97% precision by importing multimodal 
educational data. We also show how 
predictive models with merely standard 
educational data may perform statistically 
much better when combined with 
environmental, psychological, and motion 
sensor data. Since 2020, the Integrated 
Intelligent 
Intervention 
Learning 
(3I 
Learning) System in Singapore and Hong 
Kong has improved intense ABA therapy for 
more than 500 SEN kids thanks to our 
efforts.