Measuring the similarity of nodes in signed social networks with positive and negative links

Abstract

Similarity measure in non-signed social networks has been extensively studied for decades. However, how to measure the similarity of two nodes in signed social networks remains an open problem. It is challenging to incorporate both positive and negative relationships simultaneously in signed social networks due to the opposite opinions implied by them. In this paper, we study the similarity measure problem in signed social networks. We propose a basic node similarity measure that can utilize both positive and negative relations in signed social networks by comparing the immediate neighbors of two objects. Moreover, we exploit the propagation of similarity in networks. Finally, we perform extensive experimental comparison of the proposed method against existing algorithms on real data set. Our experimental results show that our method outperforms other approaches.