Comparision of several quasi-clique mining algorithms and their application to the DSS method.
Comparision of several quasi-clique mining algorithms and their application to the DSS method. This code uses the modules networkx, random, time and matplotlib.pyplot.
\ No newline at end of file
The different quasi-clique mining algorithms used are:
- Quick (Effective Pruning Techniques for Mining Quasi-Cliques 2008) implemented in the file quick.py
- Quick_redundancy_aware (same as before and adding a method from Redundancy Aware Maximal Cliques 2013) implemented in the file quick_red_aware.py
- Quick_delete_covered_edges, implemented in the file quick_del_edges.py
- A greedy quasi-clique mining algorithm implemented in greedy_qclq_mining.py
These algorithms are compared trough several aspects (runtime/compression rates/visibility), each time with and without the pre-processing algorithm (the reduction procedure of tools.py).
Three types of graph have been used:
- community graphs (generated with the file community_graph.py)
- LFR graphs (the dataset used is available here: https://zenodo.org/record/4450167?fbclid=IwAR1abb2_3qotisogJW9i_F_6RL-Cjd_czgljmh-tjONXOOzgMVfHnLWx0Ac#.Yi4FcDXjK3C and the setB folder must be placed in graph_database/LFR_graphs)
- social networks (the dataset used is available here: https://snap.stanford.edu/data/ and must be placed in graph_database/Social_networks)
To compare the algorithms run either (can take several hours):
