Learning Cellular Coverage from Real Network Configurations using GNNs
This addresses a scalability issue in self-organized networks for telecom operators, but it is incremental as it builds on existing GNN methods.
The paper tackles the problem of estimating cellular coverage quality in large areas with limited labeled data by using graph neural networks, achieving comparable accuracy to models trained on extensive labeled samples.
Cellular coverage quality estimation has been a critical task for self-organized networks. In real-world scenarios, deep-learning-powered coverage quality estimation methods cannot scale up to large areas due to little ground truth can be provided during network design & optimization. In addition they fall short in produce expressive embeddings to adequately capture the variations of the cells' configurations. To deal with this challenge, we formulate the task in a graph representation and so that we can apply state-of-the-art graph neural networks, that show exemplary performance. We propose a novel training framework that can both produce quality cell configuration embeddings for estimating multiple KPIs, while we show it is capable of generalising to large (area-wide) scenarios given very few labeled cells. We show that our framework yields comparable accuracy with models that have been trained using massively labeled samples.