A new topological metric for link prediction in directed, weighted and temporal networks

Abstract

One of the most interesting tasks in social network analysis is link prediction. There are a lot of studies dealing with link prediction task in the literature. In recent years, there is an increasing on link prediction methods trying to model network as more close to real networks such as heterogeneous, temporal and directed network models to gain better link prediction performance. Many of the existing link prediction methods don't take into account links directions in directed networks. In this paper we propose a new neighbor and graph pattern based topological metric considering direction of links for link prediction. The proposed metric also takes into account temporal and weighted information, which are useful to increase link prediction performance. Accuracy of the proposed metric is evaluated by comparison with multiple baseline metrics from literature in supervised learning methods. Experimental results demonstrate that the proposed metric improves remarkably the accuracy of link prediction.