The problem of community detection in networks is well known in the field of network science.Communities canbe described as a group of nodes with abnormally large number of edges between them as compared to edges fallingoutside the group.A quality function,"modularity",has been proposed to detect such communities (or partitions).Modularity,on maximization,is known to result in good partitions when tested on several real world networks.How-ever,modularity maximization can sometimes produce false positives I.e partitions which maximize modularity buthave no real structure.In this article,we define a "robust community structure"which can provide guarantee againstdetecting false positives by extending the notion of modularity maximization.We then develop a robust optimizationmodel to detect robust partitions of graphs which are immune to uncertainty.We present preliminary results of themodel when tested on real and synthetic data sets.Our analysis indicates that Robust Optimization provides a conve-nient framework to identify robust community structures.
展开▼
