Supplemental material for paper "Fair-mod: Fair Modular Community Detection", to be published in the proceedings for Complex Networks and their Applications 2024. Publication available here: <>.
This repository contains the implementation of the Fair-mod modularity-based community detection algorithm, with a weighted balance-based fairness. The implementation is based on the source code for Louvain community detection found in the NetworkX library, see source code here: (https://networkx.org/documentation/stable/_modules/networkx/algorithms/community/louvain.html).
The algorithm expects as input a NetworkX graph object. The graph should be undirected (directed graphs are not currently supported), and the sensitive attribute S for the graph should be coded as a node attribute named color, taking either of two values: red or blue. Future versions will address the limitations of the implementation.
The repository also includes:
process_raw.py: Code to process the raw social network datasets featured in the paper, generating the desired NX objects.s_fair_sc.py: Code for the Scalable Fair Spectral Clustering (sFairSC) algorithm, translated from the original MATLAB version of the code in https://github.com/jiiwang/scalable_fair_spectral_clustering. Credit for the algorithm goes to the original authors:
[1] Ji Wang et al. (2023). Scalable Spectral Clustering with Group Fairness Constraints. Proceedings of The 26th International Conference on Artificial Intelligence and Statistics.