This package is a very simple wrapper around hdbscan for python. It is not feature complete, but fully functional.
using HDBSCAN, Clustering
result = hdbscan(X; min_cluster_size=5, min_samples=min_cluster_size, kwargs...)
probabilities(result)
exemplars(result) # Computed at first call, this takes long time
outlier_scores(result)
result.assigments # Contains the cluster assignmentswhere X is n_features × n_points (the convention of Clustering.jl, opposite to the convention of the python library).
The result::HdbscanResult <: Clustering.ClusteringResult contains the PyObject clusterer which can be used to access everything that is not wrapped.
The label assignments are stored in result.assigments. 0 values indicate noise (-1 in python version), positive values indicate a cluster assignment.
The original documentation is available here https://hdbscan.readthedocs.io/en/latest/basic_hdbscan.html