barahona-research-group / graphbasedclustering

Multiresolution clustering of data using geometric graphs --- Code from "Graph-based data clustering via multiscale community detection" by Z Liu and M Barahona, Applied Network Science, 5 (3) (2020). See also: https://wwwf.imperial.ac.uk/~mpbara/Partition_Stability/

GraphBasedClustering

Multiscale graph-based clustering via Markov Stability

Zijing Liu

Introduction

This contains the MATLAB codes for the paper "Graph-based data clustering via multiscale community detection" by Zijing Liu and Mauricio Barahona, published in Appl Netw Sci 5, 3 (2020).

Starting from data points, described as feature vectors, the method produces different geometric graphs and applies multiscale community detection (Markov Stability) to find graph partitions at different levels of resolution, which correspond to clusterings into different numbers of clusters.

The graph-based clustering via Markov Stability uses the code in https://wwwf.imperial.ac.uk/~mpbara/Partition_Stability/ , also deposited in https://github.com/michaelschaub/PartitionStability

For an illustration, have a look at the notebook MarkovStabilityClustering.ipynb

• script_clustering_paper.m - the example file for running the framework, same as the notebook.
• test_graph_build.m - test different graph constructions.
• othercompare.m - clustering using other methods.
• matlab/ - Matlab codes for graph construction, Markov Stability and two spectral clustering methods.
• Data/ - public and generated datasets.

Citation

Liu, Z., Barahona, M. Graph-based data clustering via multiscale community detection. Appl Netw Sci 5, 3 (2020) https://doi.org/10.1007/s41109-019-0248-7