Article ID | Journal | Published Year | Pages | File Type |
---|---|---|---|---|
1888531 | Chaos, Solitons & Fractals | 2015 | 8 Pages |
Abstract
Link prediction is important for inferring interactions among members in incomplete networks. For a given snapshot of network by sparse sampling, most link prediction methods only consider one scale information, like global or local information, and it is hard to combine them together. A probabilistic model is established to give a theoretical guarantee of the information combinations. Meanwhile a bi-scale method is proposed to combine the information of microscale (neighbors) and mesoscale (communities) in the observed networks. Experiments on several social networks demonstrate that the approach always outperforms local information based methods, and it is faster than the global methods with competitive results.
Related Topics
Physical Sciences and Engineering
Physics and Astronomy
Statistical and Nonlinear Physics
Authors
Enming Dong, Jianping Li, Zheng Xie, Ning Wu,