Link Prediction Based on Information Preference Connection for Directed Network

Abstract

Link prediction aims to predict undetected or unknown links through network information. Most prediction indicators depend on local topology but ignore the pivot role of nodes. Aiming at the problem of topological sparseness in some directed scale-free networks, a preference attachment method based on Shannon theory is proposed to measure the relevance of information from the perspective of nodes. This paper defined the nodes self-information while analyzing nodes interaction process and the conditional information while analyzing the conditional probability of effective interaction based on information theory, and use them designed a new predict indicator. Experiments on 9 scale-free networks show that the proposed method has a better improvement under the Precision metrics.