Improved Variational Graph Autoencoder for Link Prediction on Real-World Telecom Graphs
Publication Date
1-1-2026
Document Type
Conference Proceeding
Publication Title
2026 1st International Conference on Emerging Trends in Advancements and Applications of Computational Intelligence Techniques Etaact 2026
DOI
10.1109/ETAACT69135.2026.11542169
Abstract
Mobility management in emerging O-RAN deployments increasingly leverages predictive intelligence to reduce handover failures and improve radio resource utilization. Recent studies have shown that Graph Neural Networks (GNNs), particularly variational graph autoencoders (VGAEs), are well suited to link-prediction formulations of user-cell association forecasting. In this work, next-cell selection is cast as a link-prediction problem on a bipartite user-cell graph derived from a real-world Huawei RW deployment and enriched with edge features. The proposed attention-based VGAE achieves an accuracy of 0.9127 and an F1-score of 0.9190 on the 10k-UE RW graph, improving over a baseline accuracy of 0.80 and F1-score of 0.83, and thereby providing substantially more reliable next-cell predictions for proactive handover.
Keywords
Graph Neural Networks, Link Prediction, O-RAN, Telecom Graphs, VGAE
Department
Computer Science
Recommended Citation
Mugdha Arun Sathe, Pooja Shyamsundar, Ojas Naik, and Navrati Saxena. "Improved Variational Graph Autoencoder for Link Prediction on Real-World Telecom Graphs" 2026 1st International Conference on Emerging Trends in Advancements and Applications of Computational Intelligence Techniques Etaact 2026 (2026). https://doi.org/10.1109/ETAACT69135.2026.11542169
