CO-clustering for STAR-structured data (CoStar) is a multi-view co-clustering algorithm. Given a set of different views of the same dataset CoStar provides, at the same time, a clustering on documents and a clustering on features of each view.
This implementation is compatible with the scikit-learn (http://scikit-learn.org/) clustering interface.
If using, cite the following paper:
[1] Ienco, Dino, et al., 2013. Parameter-less co-clustering for star-structured heterogeneous data. Data Mining and Knowledge Discovery 26.2: 217-254
To use the algorithm, follow these simple steps:
-
Install project dependencies with pip
pip install -r requirements.txt -
Test the code running the file
costar-test.pypython costar-test.py
Future works are related to:
- Improve the algorithm performances
- Improve code parallelization
Python implementation by Valentina Rho.
If you want to contribute or comment, write to pensa@di.unito.it.
The software is released under GPL v3 licence.