Adaptive Learning Systems Based on Deep Learning for the Diagnosis and Support of Learning Disabilities

Abstract

In the current field of educational technology, adaptive learning systems have become a key tool in supporting personalized learning. Particularly for students facing learning disabilities, how to provide effective diagnosis and support has become an important topic. Learning disabilities often hide within the behavioral characteristics of learners, requiring detailed analysis for identification and intervention. This study aims to leverage deep learning technology, especially the XGBoost algorithm, to improve the accuracy of diagnosing learning disabilities. It also seeks to implement knowledge transfer across different subjects through transfer learning algorithms, thereby overcoming learning disabilities. Research background indicates that despite the increasing application of adaptive learning systems, existing methods for diagnosing and supporting learning disabilities still have limitations. These systems often fail to accurately parse complex behaviors and cognitive patterns of students, resulting in insufficiently personalized assistance. Moreover, students with learning disabilities often encounter more difficulties in interdisciplinary learning, and existing learning systems have not effectively supported their cross-domain knowledge transfer. The diagnostic method based on XGBoost and the adaptation approach through transfer learning proposed in this paper have been experimentally verified to effectively improve the prediction accuracy of learning disabilities and promote knowledge transfer between different subjects for students, helping them overcome learning disabilities in specific disciplines. This research not only has significant implications for enhancing the intelligence level of adaptive learning systems but also positively impacts achieving educational equity and enhancing the overall learning experience of learners.