Link prediction in complex networks via modularity-based belief propagation<xref rid="cpb_26_3_038902_fn1" ref-type="fn">*</xref> <fn id="cpb_26_3_038902_fn1"> <label>*</label> <p>Project supported by the National Natural Science Foundation of China (Grants No. 61202262), the Natural Science Foundation of Jiangsu Province, China (Grants No. BK2012328), and the Specialized Research Fund for the Doctoral Program of Higher Education of China (Grants No. 20120092120034).</p> </fn>

Link prediction in complex networks via modularity-based belief propagation* *

Project supported by the National Natural Science Foundation of China (Grants No. 61202262), the Natural Science Foundation of Jiangsu Province, China (Grants No. BK2012328), and the Specialized Research Fund for the Doctoral Program of Higher Education of China (Grants No. 20120092120034).

Lai Darong^{1, 2, †}, Shu Xin³, Nardini Christine^{4, 5}

(color online) Comparisons on the performance of link prediction in synthetic networks generated by the stochastic block model with nodes, the number of blocks q = 10, and different average degrees c. BMPA was compared with RA, AA, CN, and PA. Performance of link prediction on networks with c = 3 ((a)–(c)), c = 6 ((d)–(f)), and c = 10 ((g)–(i)). For missing link prediction (left panel), a fraction of links were randomly removed from the network but the networks were kept connected, while for evolving and spurious link prediction (middle and right panel) a fraction of links were randomly added on the networks. Each point is averaged over 10 network instances by repeating the experiments 30 times.