HCDF: A Hybrid Community Discovery Framework

摘要: We introduce a novel Bayesian framework for hybrid community discovery in graphs. Our framework,HCDF (short Hybrid Community Discovery Framework), can effectively incorporate hints from number of other detection algorithms and produce results that outperform the constituent parts. describe two HCDF-based approaches which are: (1) effective, terms link prediction performance robustness to small perturbations network structure; (2) consistent, effectiveness across various application domains; (3) scalable very large graphs; (4) nonparametric. extensive evaluation on collection diverse real-world graphs, with millions links, show our (a) achieve up 0.22 improvement as measured by area under ROC curve (AUC), (b) never have an AUC drops below 0.91 worst case, (c) find communities are robust structure defined Variation Information (an entropybased distance metric).