HiReCS is a C++ clustering library for the multi-scale hierarchical community structure discovery with crisp overlaps in large evolving networks having nearly linear computational complexity.
See details in http://www.lumais.com/hirecs, documentation and datasets in http://goo.gl/C4gHH0
API Manual: http://www.lumais.com/doxy/hirecs/index.html
Releases: https://github.com/XI-lab/hirecs/releases
- WARNING
The library is under development now, it currently has issues with datasets processing of the size more than 100 000 links.
Currently it forms excellent root level (top-level, coarse-grained) clusters, but might have issues with the underlying levels of the hierarchy.
If you find any issues, please submit them here.
$ ./hirecs -oje dataovp/pentagon.hig
High Resolution Hierarchical Clustering with Stable State (HiReCS) started, nodes: 5, weight: 10
Initial state, mod: -0.2, nodes num: 5
Initial nodes clustering is completed
Folding is completed, mod: 0.05, dmod: 0.25, weight (both dir): 10, clusters num: 5
Folding is completed, mod: 0.175, dmod: 0.125, weight (both dir): 10, clusters: 5 (cls: 5, prop: 0) / 10, dcls: 0
Hierarchy building is completed, mod: 0.175, weight (both dir): 10, clusters: 10, root size: 5
-Root size: 5
{"root":[5,6,7,8,9],"clusters":{"0":{"owners":[5,6],"des":[0,1],"leafs":true},"1":{"owners":[5,7],"des":[0,2],"leafs":true},"2":{"owners":[6,8],"des":[1,3],"leafs":true},"3":{"owners":[7,9],"des":[2,4],"leafs":true},"4":{"owners":[8,9],"des":[3,4],"leafs":true},"5":{"des":[0,1]},"6":{"des":[0,2]},"7":{"des":[1,3]},"8":{"des":[2,4]},"9":{"des":[3,4]}},"communities":{"5":{"0":0.5,"2":0.25,"1":0.25},"6":{"1":0.5,"3":0.25,"0":0.25},"7":{"2":0.5,"4":0.25,"0":0.25},"8":{"3":0.5,"4":0.25,"1":0.25},"9":{"3":0.25,"4":0.5,"2":0.25}},"nodes":5,"mod":0.175}
Note: the binaries are located in bin/Release/, target OS: Linux Ubuntu 14.04 x64 (ask me if you need another target architecture, the code is crossplatform).
- HiCBeM - Benchmark for the Hierarchical Clustering Algorithms: https://github.com/XI-lab/hicbem
If you are interested in this clustering library, please visit eXascale Infolab where you can find another projects and research papers related to Big Data!