Intelligent Diagnostics and Self-Healing of Beyond-5G Networks: Critical Analysis, Evolution Roadmap, and Educational Module Based on the 5G3E Dataset

This research presents a comprehensive 18-part critical analysis and evolution roadmap for the 5G3E dataset — the first open multi-layer benchmark for Beyond-5G network automation, created at CNAM (Paris, 2022) by Phung, Yellas, Bin Ruba, and Secci. Through systematic forensic analysis, we identify 31 critical limitations spanning architectural, statistical, and methodological dimensions, including: absence of 5G Standalone and Network Slicing; idealized radio channel without fading, mobility, or handover; primitive anomaly injection mechanisms (stress-ng, iptables) incompatible with 5G physics; 45% of operator trace fields unused; 80 ms sampling causing aliasing of RAN events; Forward-Fill imputation generating 3.4M artificial zero-derivative artifacts; and fundamental incompatibility with 3GPP Release 16/17 and O-RAN architectures. We propose 5G3E v2.0 — a complete redesign specification featuring: full 5G SA core (AMF/SMF/UPF/PCF/NSSF), three standard slices (eMBB/URLLC/mMTC) with 5QI enforcement, realistic RAN channel models (Rayleigh fading, 3D ray-tracing, SUMO mobility), 30+ annotated anomaly types with causal graphs and XAI labels, adaptive sampling (10-100 ms), Kalman-filter-based imputation, and an official ML benchmark protocol with 10+ baseline models. The work includes: working Python code (PyTorch LSTM Autoencoder + SHAP + NetworkX causal graphs), a 16-week educational module with 4 laboratory exercises, peer review simulation, budget estimation (EUR 260K-370K), and a 4-year PhD research roadmap targeting 8-12 publications. Original dataset reference:Phung, C. D., Yellas, N.-E.-H., Bin Ruba, S., & Secci, S. (2022).An Open Dataset for Beyond-5G Data-driven Network Automation Experiments.In 2022 1st International Conference on 6G Networking (6GNet), Paris.DOI: 10.1109/6GNet54646.2022.9830292. HAL: hal-03698732v2.