https://doi.org/10.1051/epjconf/202532801072
Design of an Improved Model for Personalized Adaptive E-Learning Using Context-Aware Federated Learning and Hierarchical Semantic Graph Analysis
1 Computer Science & Engg., Gandhi institute of engineering and technology University, Gunupur, Odisha, India
2 CSE Dept, Gandhi Institute of Engineering and Technology University, Gunupur, India
3 Aditya Institute of technology and management, Tekkali, India
* Corresponding author: sirishaphd123@gmail.com
Published online: 18 June 2025
The improvement of e-learning platforms has raised a challenge of building an adaptive, personalized, and privacy-preserving scalable system, which can service hugely varied types of learners. The challenges faced by most traditionally adaptive systems include limited personalization, context unawareness, low explanation, and reliance on shallow metrics of engagement. Most existing federated learning models left out the real-life context of the learners while semantic understanding modules do less in multi-level cognitive states capturing. The current reinforcement learning techniques in education are centered around the short-term task optimization whereas immersive feedback mechanisms do not usually extend beyond simple audiovisual interactions in process. This work presents IMEPAL: an Iterative Multi-Modal Personalized Adaptive Learning framework developing five new ways to overcome these challenges. Context-Aware Federated Personalization Learning (CAFPL) defines dynamic, private personalizedness by embedding the learner's context into federated updates. Hierarchical Semantic Graph Attention Networks (H-SGAN) facilitate a fine-grained semantic-cognitive analysis, identifying both topic relations and student cognitive conditions. Dynamic Reinforcement Meta-Learning Optimization (DRMLO) dynamically structures rewards based on evolved skill graphs to strengthen long-term learning. Multi-Sensory Augmented Virtual Feedback Learning (MSAVFL) encourages real-time, multi-sensory interaction in a VR/AR environment for improved learning. Lastly, Neuro-Symbolic Personalized Knowledge Reasoning (NSPKR) provides explainable, logically sound recommendations for learning through merging neural embeddings with symbolic rule-based reasoning process. Preliminary results demonstrated significant improvements: +15-20% in personalization robustness, +18% in semantic understanding, +20-23% in learning retention, an emotional engagement normalized score of approximately 0.89, and a +19% increase in explainability acceptance in process. Thus, IMEPAL presents an all-around adaptive learning solution, privacy protected and explainable transformation for the future of intelligent eLearning systems.
© The Authors, published by EDP Sciences, 2025
This is an Open Access article distributed under the terms of the Creative Commons Attribution License 4.0, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
