Measuring node influence in software networks is critical for identifying key functional components. However, existing methods rely on limited structural features and lack robust validation. We propose a novel node ranking approach(CLG) that integrates structural hole theory to capture network Constraint, defect aggregation and propagation analysis to represent Local fault impact, and length-penalized shortest path centrality to assess Global information dissemination. Validation employs a dual framework: Ranking capability metrics (distinctness, monotonicity) assess node discrimination and ranking consistency; Correctness verification combines SIR epidemic simulations with groundtruth correlation analyses. Experiments on five open-source software networks show our method consistently outperforms 14 baseline approaches across key metrics.
Yu et al. (Thu,) studied this question.