An improvement of link prediction by combining local information and betweenness

Link prediction has significance in both theoretical interest and practical operation. Many methods via local and global structural information have been proposed. Methods based on local information like the Common Neighbours Index(CN) successfully reduce the computational expense but suffer from poor prediction performance. In this article, we put forward a new approach, namely Betweenness and Common Neighbours Index(B-CN) which combines the betweenness and the CN index. Comparing the new method with the CN index in ten real networks, the numerical results indicate that the approach has a great improvement than the CN index, especially when the fraction of missing links is great.