Background

A very large number of community detection techniques have been developed. We survey some of these, focusing on those that use both structure (the explicit similarities from relationships) and attributes (the similarities based on common interests). Most techniques look for disjoint communities, but it is also plausible to allow communities to overlap, that is to allow objects to belong to more than one community. Community detection techniques range from those that define a community quality measure and then find communities that optimise this measure, to hybrid techniques that use a wide variety of algorithmic pieces including, recently, deep learning. The literature is large and disparate, so we concentrate on significant examples.