Research on Link Prediction Algorithm Integrating High-order Information and Node Centrality in Network

Link prediction is an important branch in the field of complex network research. Traditional similarity-based link prediction algorithms mainly calculate the similarity between nodes based on single features such as common neighbors and degrees, without considering the impact of higher-order network structures on link formation. To address this issue, this paper proposes a link prediction algorithm called (H-CCNC) and its extended version (FH-CCNC), which integrates high-order information and node centrality by considering node degree, node centrality, and high-order clustering coefficient from both local and global perspectives in the network. Experiments were conducted on real datasets with different densities, using the area under the curve (AUC) and precision as evaluation metrics and comparing with traditional link prediction algorithms. The results show that H-CCNC and FH-CCNC can predict missing links in the network more accurately, with an AUC improvement ranging from 1% to 22% on different datasets.