Amazon DenseClus
DenseClus is a Python module for clustering mixed type data using UMAP and HDBSCAN. Allowing for both categorical and numerical data, DenseClus makes it possible to incorporate all features in clustering.
Installation
python3 -m pip install Amazon-DenseClus
Usage
DenseClus requires a Panda's dataframe as input with both numerical and categorical columns.
All preprocessing and extraction are done under the hood, just call fit and then retrieve the clusters!
from denseclus import DenseClus
clf = DenseClus(
umap_combine_method="intersection_union_mapper",
)
clf.fit(df)
print(clf.score())
Examples
A hands-on example with an overview of how to use is currently available in the form of a Jupyter Notebook.
References
@article{mcinnes2018umap-software,
title={UMAP: Uniform Manifold Approximation and Projection},
author={McInnes, Leland and Healy, John and Saul, Nathaniel and Grossberger, Lukas},
journal={The Journal of Open Source Software},
volume={3},
number={29},
pages={861},
year={2018}
}
@article{mcinnes2017hdbscan,
title={hdbscan: Hierarchical density based clustering},
author={McInnes, Leland and Healy, John and Astels, Steve},
journal={The Journal of Open Source Software},
volume={2},
number={11},
pages={205},
year={2017}
}
