Approximating Network Centrality Measures Using Node Embedding and Machine Learning

Extracting information from real-world large networks is a key challenge nowadays. For instance, computing node centrality may become unfeasible depending on the intended due to its computational cost. One solution develop fast methods capable of approximating network centralities. Here, we propose an approach for efficiently centralities using Neural Networks and Graph Embedding techniques. Our proposed model, entitled Network Centrality Approximation (NCA-GE), uses adjacency matrix graph set features each (here, use only degree) as input computes approximate desired rank every node. NCA-GE has time complexity O(|E|), E being edges graph, making it suitable networks. also trains pretty fast, requiring thousand small synthetic scale-free graphs (ranging 100 1000 nodes each), works well different centralities, sizes, topologies. Finally, compare our state-of-the-art method that approximates ranks degree eigenvector input, where show outperforms former in variety scenarios.