Mechanism-Embedded Machine Learning-Driven Resilience Governance of Digital–Intelligent Innovation Collaboration Networks

The existing research on digital–intelligent innovation collaboration networks has not effectively connected the evolution of real collaborative relationships with governance methods, and governance models often struggle to accommodate both empirically identified evolutionary mechanisms and sequential decision-making capability. Using Chinese digital–intelligent innovation co-patent data from 2016 to 2025, this study develops a machine learning analytical framework centered on relationship evolution and collaboration resilience governance and employs OSR-DDQN to optimize dynamic intervention strategies under limited budgets. The results show that state transitions in digital–intelligent innovation collaboration exhibit pronounced path specificity, with the innovation capability, relational capital, and technological complementarity displaying differentiated conditional associations across transition paths. Further simulation experiments show that, across different budget conditions, OSR-DDQN improves the mean reward by 11.56% to 12.83% relative to the best-performing standard deep Q-network. Within the specified governance environment, once the budget reaches a minimally sufficient level, further performance gains arise primarily from optimizing the action timing, intervention targets, and stopping decisions rather than from simply increasing governance resources.