Research on Federated Transfer Learning Algorithms for Real-Time Analysis of Students' Learning Situation and Personalized Paths under Edge Computing Architecture

Abstract

Against the backdrop of educational digitalization, traditional cloud-based learning situation analysis faces challenges such as insufficient real-time performance, data privacy risks, and difficulties in cross-domain adaptation. This study proposes a three-level algorithm system and constructs a collaborative architecture consisting of edge preprocessing, federated optimization, and transfer adaptation.The edge layer realizes lightweight feature extraction of multimodal data; the federated layer adopts the FedProx algorithm combined with differential privacy and homomorphic encryption to ensure the security of cross-school collaboration; the transfer layer uses domain adversarial training and knowledge graph reinforcement learning to achieve cross-domain feature alignment and personalized learning path generation. Experiments show that the algorithm achieves low-latency inference on the edge side, effectively improves the accuracy of cross-domain recommendations, and reduces privacy risks, providing a practical technical solution for educational intelligence.