Digging Deeper: Exploring Extended Common Neighbors in Link Prediction

Link prediction is a prominent area of research in complex network analysis, with applications in various fields including network evolution analysis, recommendation systems, and protein interaction analysis in biological networks. The main objective of link prediction is to estimate missing or hidden links between disconnected nodes in a network. Over the years, numerous algorithms and models have been developed for link prediction, categorized into similarity-based approaches and learning-based approaches.

Similarity-based approaches calculate similarity scores between unconnected nodes based on available information. The underlying hypothesis is that nodes with higher similarity are more likely to have a link between them. This idea is straightforward and widely explored, making similarity-based approaches the mainstream in link prediction research. The Common Neighbors (CN) index, for instance, simply counts the number of common neighbors between two nodes. The Adamic-Adar (AA) and Resource Allocation (RA) indexes, variations of the CN index, adjust for the influence of highly connected common neighbors. These indexes are considered local methods as they rely on local structural information.

In addition to local methods, global and quasi local methods have been proposed by researchers, such as Katz, SimRank, Random Walks with Restart, Local Path, FriendLink, and Local Random Walk. However, as complex networks continue to grow in size, local methods remain…