Predicting missing links via local feature of common neighbors

Abstract

Predicting missing links in complex networks has a theoretical interest and practical significance in social network analysis. In this paper, we study the link prediction results as the change of the exponent on common neighbor's degree and find some regular pattern between different networks and different exponent. Then we present a local network metric based on the regular pattern to estimate the likelihood of the existence of a link between two nodes. Our new metric takes the exponent of common neighbor's degree into consideration for link prediction. The paper also recommends the value range of the exponent. We compare nine well-known local information metrics and the new metric on eight real networks. The result evaluated by AUC indicates that our new metric, namely Degree Exponent Change metric, have a better prediction accuracy than other well-known metrics.