The hdbscan Clustering Library

The hdbscan library is a suite of tools to use unsupervised learning to find clusters, or dense regions, of a dataset. The primary algorithm is HDBSCAN* as proposed by Campello, Moulavi, and Sander. The library provides a high performance implementation of this algorithm, along with tools for analysing the resulting clustering.

User Guide / Tutorial

• Basic Usage of HDBSCAN* for Clustering
  • The Simple Case
  • What about different metrics?
  • Distance matrices
• Getting More Information About a Clustering
  • Condensed Trees
  • Single Linkage Trees
• Parameter Selection for HDBSCAN*
  • Selecting min_cluster_size
  • Selecting min_samples
  • Selecting alpha
• Outlier Detection

Background on Clustering with HDBSCAN

• How HDBSCAN Works
  • Transform the space
  • Build the minimum spanning tree
  • Build the cluster hierarchy
  • Condense the cluster tree
  • Extract the clusters
• Comparing Python Clustering Algorithms
  • Some rules for EDA clustering
  • Getting set up
  • Testing Clustering Algorithms
  • K-Means
  • Affinity Propagation
  • Mean Shift
  • Spectral Clustering
  • Agglomerative Clustering
  • DBSCAN
  • HDBSCAN
• Benchmarking Performance and Scaling of Python Clustering Algorithms
  • Comparison of all ten implementations
  • Comparison of fast implementations
  • Comparison of high performance implementations
  • But should I get a coffee?
  • Conclusions

API Reference

• API Reference
  • HDBSCAN
  • RobustSingleLinkage

Indices and tables

• Index
• Module Index
• Search Page