ZERO NODES EFFECT: VALID LINK PREDICTION IN SPARSE NETWORKS

Abstract

Zero-degree nodes are an important difficulty in sparse networks' link prediction. Clustering and preferential attachment, as the most important characteristics of complex networks, have been paid little attention in similarity indices. Inspired by the coexistence of clustering and preferential attachment in real networks, this paper proposes a new preferential attachment index and new clustering index, which have here been integrated into a hybrid index that considers the dynamic evolutionary forces of complex networks and can solve the problem of excessive zero-degree nodes in sparse networks and check evolution mechanism. Experiments proved prediction accuracy can be remarkably enhanced.