TeleGraph: A Benchmark Dataset for Hierarchical Link Prediction

TL;DR

TeleGraph addresses the problem of link prediction in highly hierarchical and sparse networks by introducing a real-world telecommunication dataset with rich alarm-based node attributes. The paper benchmarks a spectrum of methods—from heuristic measures to embedding-based models and SEAL-style subgraph classifiers—demonstrating that many traditional approaches fail on near-tree-like graphs, while subgraph-focused GNNs and one-hot alarm features offer the strongest performance. Key contributions include public release of TeleGraph (41,143 nodes, 41,424 edges, 240 alarm types), a descriptive and exploratory analysis of its tree-like structure, and comprehensive benchmark results across multiple baselines using $AUC$ and $AP$. The dataset provides a challenging, realistic benchmark to foster development of topology- and attribute-aware link prediction methods with practical relevance for fault management in large-scale telecommunication networks.

Abstract

Link prediction is a key problem for network-structured data, attracting considerable research efforts owing to its diverse applications. The current link prediction methods focus on general networks and are overly dependent on either the closed triangular structure of networks or node attributes. Their performance on sparse or highly hierarchical networks has not been well studied. On the other hand, the available tree-like benchmark datasets are either simulated, with limited node information, or small in scale. To bridge this gap, we present a new benchmark dataset TeleGraph, a highly sparse and hierarchical telecommunication network associated with rich node attributes, for assessing and fostering the link inference techniques. Our empirical results suggest that most of the algorithms fail to produce a satisfactory performance on a nearly tree-like dataset, which calls for special attention when designing or deploying the link prediction algorithm in practice.

Figures (6)

• Figure 1: An illustration of the access network which is the most basic facility in the telecommunication network. It is responsible for connecting subscribers to the immediate service provider via base stations, transmission, and route. The overall topology is tree-like with a distinct hierarchical layout, but some local areas are grid or circular. Recovering the missing topology between the devices enables the efficient fault management fournier2020discovering, which is of importance to ensure the stability and reliability of the network.
• Figure 2: The network topology (left) with alarm attributes (right) of the TeleGraph dataset.
• Figure 3: Number of different device types
• Figure 4: Degree distribution of TeleGraph, which is asymptotically power-law distributed
• Figure 5: Node number distribution of the cycles