Link prediction from fusion information

Link prediction of complex network aims at predicting the likelihood of the existence of unknown links. In the most previous works, many kinds of information, such as common neighbor, local path, transition probability, have been used to predict the unknown links. However, methods using fusion information are relatively rare. In this paper, different kinds of information are fused with community information and a method using fusion information is proposed. According to the objective function of density of modularity in the method for community detection, an index based on community information is defined and is further fused with the existing indexes based on transition probability and degrees of nodes, to predict the unknown links. Indexes based on fusion information are compared with not only frequently-used, basic indexes, but other methods using fusion information. Experimental results show that indexes based on fusion information perform efficiently and accurately. & COPY; 2023 Published by Elsevier B.V.

Automated summary

In this paper, a method using fusion information is proposed to predict the likelihood of unknown links in complex networks. By combining community information with traditional indexes based on transition probability and node degrees, the proposed method outperforms frequently-used basic indexes and other fusion information methods in link prediction performance.