In order to overcome the shortcomings of the existing key node recognition technologies， such as high computational complexity， single evaluation dimension， and limited application scope， this paper constructs a novel algorithm suitable for key node evaluation. This algorithm first evaluates the local influence of a node by analyzing its information entropy and the influence contribution of its neighboring nodes， thereby eliminating the shortcomings of traditional evaluation criteria that rely solely on node metrics. Secondly， this algorithm determines the global influence of nodes by measuring the correlation of distances between them， effectively solving the problem of excessive computation caused by considering multiple paths. In order to fully demonstrate the practicality of the algorithm， four real networks of different scales and six comparative algorithms are analyzed using monotonicity experiments， infectious disease model experiments， and robustness experiments. The final results show that the algorithm has certain improvements in accuracy， effectiveness， and recognition ability. At the same time， its computational complexity is low and it can be applied to sparse networks.