The Shared Nearest Neighbor clustering algorithm [1], also known as SNN, is an extension of DBSCAN that aims to overcome its limitation of not being able to correctly create clusters of different densities. This work provides a Python 3 implementation for SNN following the conventions of the scikit-learn library, and compares its results to multiple datasets with other, more traditional clustering methods.
[1] Ertöz, L., Steinbach, M., & Kumar, V. (2003, May). Finding clusters of different sizes, shapes, and densities in noisy, high dimensional data. In Proceedings of the 2003 SIAM international conference on data mining (pp. 47-58). Society for Industrial and Applied Mathematics.
| Author | Albert Espín |
| Date | April 2019 |
| Code icense | MIT |
| Report license | Creative Commons Attribution, Non-Commercial, Non-Derivative |
