Link prediction in complex networks based on the interactions among paths

Abstract

Link prediction in incomplete complex networks is an important issue in network science. Recently, various structure-based similarity methods have been proposed. However, most path-dependent methods merely pay attention to the contributions of paths with specific length, which neglects the interactions of paths with different length for performance improvement. Motivated by the resource-traffic flow mechanism on networks, we measure the interaction relationship of paths with a resource receiving process. In this process, each node takes certain initial resources quantified by its H-index, and then the intermediate nodes on paths can receive resources from their neighbours. Based on this process, a local path-based link predictor which emphasizes the effect of the Resources from Short Paths (RSP) is proposed. Experiments on twelve real-world networks demonstrate that the RSP index has better performance than other nine structure-based similarity methods.