Comparison of group discovery methods on datasets with ground-truth

Abstract

Finding social network communities (groups) is fundamental in understanding the properties of the whole network and better understanding human behavior. Many definitions of such structures have been proposed, and therefore, also a lot of algorithms for finding them. These algorithms differ in many aspects and each of them has additionally a number of parameters that need to be set apriori. The article presents a comparison of the results of using different algorithms for datasets that have a ground truth. Moreover, for nondeterministic algorithms, the variability of their results was also analyzed.