Mining key nodes in the network plays a great role in the evolution of information dissemination, virus marketing, and public opinion control, etc.The identification of key nodes can effectively help to control network attacks, detect financial risks, suppress the spread of viruses diseases and rumors, and prevent terrorist attacks.In order to break through the limitations of existing node influence assessment methods with high algorithmic complexity and low accuracy, as well as one-sided perspective of assessing the intrinsic action mechanism of evaluation metrics, a comprehensive influence (CI) assessment algorithm for identifying critical nodes was proposed, which simultaneously processes the local and global topology of the network to perform node importance.The global attributes in the algorithm consider the information entropy of neighboring nodes and the shortest distance nodes between nodes to represent the local attributes of nodes, and the weight ratio of global and local attributes was adjusted by a parameter.By using the SIR (susceptible infected recovered) model and Kendall correlation coefficient as evaluation criteria, experimental analysis on real-world networks of different scales shows that the proposed method is superior to some well-known heuristic algorithms such as betweenness centrality (BC), closeness centrality (CC), gravity index centrality(GIC), and global structure model (GSM), and has better ranking monotonicity, more stable metric results, more adaptable to network topologies, and is applicable to most of the real networks with different structure of real networks.