A Comparison of Objective Functions in Network Community Detection

Community detection, as an important unsupervised learning problem in social network analysis, has attracted great interests in various research areas. Many objective functions for community detection that can capture the intuition of communities have been introduced from different research fields. Based on the classical single objective optimization framework, this paper compares a variety of these objective functions and explores the characteristics of communities they can identify. Experiments show most objective functions have the resolution limit and their communities structure have many different characteristics.