... integrates traditional network centrality analysis with graph neural networks to capture both structural influence and flow mediation. Using operational schedule data, Bonacich power and random-walk betweenness centrality were embedded into a graph learning model to evaluate node importance beyond...