This repository contains some software used in Scalable Motif-Aware Graph Clustering. You can find the project web page here.

• Charalampos E. Tsourakakis
• Jakub Pachocki
• Michael Mitzemacher

Our code takes as input a graph, and the ground-truth communities, check files com-amazon.ungraph.txt and com-amazon.top5000.cmty.txt to see the input format.

1. Run from a terminal python relabel-graph.py com-amazon.ungraph.txt com-amazon.top5000.cmty.txt amazon.mace amazon.communities.
2. Download and compile MACE.
3. Run ./mace C -l 3 -u 3 amazon.mace amazon.triangles
4. Run python mace-to-list.py amazon.mace amazon.edges
5. For some stats, run python community-stats.py amazon.edges amazon.communities amazon.edges.stats and python community-stats.py amazon.triangles amazon.communities amazon.triangles.stats
6. Compile the file triangle_clusters.cpp g++ -std=c++11 -o triangle_clusters triangle-clusters.cpp
7. Run ./triangle_clusters amazon.triangles numbers_of_nodes (334858 for amazon) threshold_value > amazon_clusters.txt
8. Quick heuristic evaluation python grade-clusters.py amazon.communities amazon_clusters.txt output.txt