Identifying Influential Nodes in Social Networks by Integrating the Centrality Method and Node Activity

Nowadays, social networks have become a strong tool among researchers in addition to their social functions. This tool has many applications in identifying crimes, criminals and terrorists, solving epidemic problems, successful marketing and other topics in various fields. The researchers are using the influence maximization (IM) to achieve these goals. The task of maximization is to identify the influential nodes that are known as the seed nodes. It is a  strategy to achieve the maximum information diffusion or minimum epidemy with minimal cost. Since maximization is an NP-hard problem, researchers are looking for ways to reduce the complexity and acceptable identification accuracy by identifying influential nodes. Therefore, to overcome the complexity and increase the identification  accuracy, in this research a new method with activity-centrality combination is proposed. In this approach, to extract nodes by the centrality method a total constraint is constructed on the network graph in order to proceed to the local nodes extracted from the node activity analysis. The results of analyzing the activity of each node are combined with its    centrality method score which ultimately leads to the identification of influential nodes. The proposed method is compared with other methods such as PageRank and Closeness Centrality methods, and the evaluation results show that whilst having a lower complexity, the proposed method is better than both in terms of accuracy. In the future, the concepts of repetitive scoring can be used to further enhance the accuracy of the activity analysis phase.