Node importance ranking for influence maximization in social networks

With the rapid advancement of network technology, social networks have become important tools for observing and understanding the world. In social network research, node importance ranking and influence maximization have gained significant attention due to their applications in public opinion control and marketing. Existing ranking methods often rely on single network features, overlooking the combined effects of multiple attributes on node influence, leading to poor performance in complex scenarios. Traditional influence maximization approaches select the top k nodes as seed sets based on importance rankings, but their joint influence does not equate to the simple sum of individual seed influences, resulting in an overlap issue. To address these problems, this paper proposes a multi-attributes node importance ranking algorithm based on entropy and analytic hierarchy process, which integrates degree centrality, K-shell values based on location, and PageRank values from random walks. Additionally, to accurately capture the influence scope of nodes, a high-frequency influence subnet is constructed, which forms a new network from frequently activated nodes and edges. Based on the high-frequency influence subnet, two improved dynamic deduplication-based algorithms are introduced to reducing influence overlap among seed nodes. Experimental results show that these improved algorithms achieve a broader distribution of seed nodes and greater influence spread, effectively resolving the overlap problem.