The rapid growth of 5G networks an‎d the expected arrival of 6G make it hard to maintain mobile servicesʹ excellent quality of experience (QoE). Latency an‎d scalability problems limit traditional cloud-based QoE prediction frameworks, rendering them insufficient for real-time, user-centered applications. This paper presents an Edge AI-driven system that leverages Radio Access Network (RAN) data, including SINR, RSRP, an‎d han‎dover information, to conduct real-time QoE prediction at the network edge. The system allows for low-latency, energy efficient inference in resource-limited settings using lightweight ensemble models like Ran‎dom Forest an‎d LightGBM. The study also studies federated an‎d personalized learning strategies to boost model adaptability while respecting data privacy. A complete simulation employing NS 3 an‎d real-world RAN data will test model accuracy, inference speed, an‎d energy efficiency. The projected conclusion is a scalable, privacy-preserving technology that enables proactive QoE optimization in varied deployment settings. This helps make AI-native, self-organizing 6G networks a reality, helps network operators keep more users, an‎d improves service delivery by using innovative edge-based resource management.

6/10/2025 12:00:00 AM

86952

1404/04/23

Edge AI-Based User Experience Prediction in Mobile Networks Using RAN Data Analytics

Edge Artificial Intelligence , Quality of Experience (QoE) , RAN Telemetry , 5G/6G Networks , Machine Learning , Federated Learning , Real-time Inference , Energy-efficient Models