Intelligent Recommendation for Personalized Educational Intervention for Children with Autism

Educational intervention is one of the most important and effective treatment methods for children with autism spectrum disorder (ASD). However, limited educational resources are unable to meet the increasing needs of ASD patients. To this end, we propose a recommendation algorithm based on graph neural network (GNN), which intelligently recommends intervention activities and games for personalized needs of children with ASD. The proposed method characterizes ASD users according to their PEP-3 (Psycho educational Profile, PEP) scale evaluation data. Since the data on children's intervention behaviour contains multiple types of entity relationships, we construct graph model through meta-paths to mine deeper semantic associations and enable recommendations for intervention activities and games. The experimental results validate that our method perform well in all four recommended evaluation indicators, proving its effectiveness.