(This site has been archived. Visit a newer version here.)
Welcome to the website for the Community Structure Analysis Framework! The tools on this site provide researchers and practitioners with a novel, rigorous, machine learning-based method for understanding and comparing the structures of communities produced by different algorithms. It can be used to identify pairs of algorithms that produce similar types of communities, and identify algoritihms that produce especially distinctive communities. When a user has examples of "real" communities, this framework can guide selection an appropriate community detection algorithm to locate other similar communities.
To use this method, one first collects a set of community detection algorithms (examples can be found here), and by applying each of these algorithms to the network under study, obtains examples of communities from different algorithm classes. For each community, one then calculates a feature vector measuring important structural characteristics of that set. Then by using classifiers such as a Support Vector Machine (SVM), one can identify similarities and differences between, as well as understand nuances of, each type of community.
Welcome to the website for the Community Structure Analysis Framework! The tools on this site provide researchers and practitioners with a novel, rigorous, machine learning-based method for understanding and comparing the structures of communities produced by different algorithms. It can be used to identify pairs of algorithms that produce similar types of communities, and identify algoritihms that produce especially distinctive communities. When a user has examples of "real" communities, this framework can guide selection an appropriate community detection algorithm to locate other similar communities.
To use this method, one first collects a set of community detection algorithms (examples can be found here), and by applying each of these algorithms to the network under study, obtains examples of communities from different algorithm classes. For each community, one then calculates a feature vector measuring important structural characteristics of that set. Then by using classifiers such as a Support Vector Machine (SVM), one can identify similarities and differences between, as well as understand nuances of, each type of community.
Watch our KDD Madness video:
See our KDD slides:
