Node Importance Assessment Method Based on Double Deep Q-Network

The assessment of node importance in complex networks is a current research focus in network science. The accuracy of such evaluations has a direct impact on the security and development of real-world complex networks in different sectors. This paper presents a method for assessing the importance of nodes, Dws. It combines the double deep Q-Network as a core model with entropy weighting and reordering techniques. Comparing Dws with five classical methods for importance identification, two simulated experiments are conducted using infectious disease models for node ranking and key node identification in six real-world complex networks. The strong discriminative ability, robustness, and applicability of Dws in appraising the ranking of node importance and pinpointing crucial nodes is substantiated by the results.