hdbscan-ts

A TypeScript implementation of HDBSCAN (Hierarchical Density-Based Spatial Clustering of Applications with Noise), based on Campello et al. 2017.

HDBSCAN is particularly effective at:

• Detecting clusters of varying densities
• Identifying noise points
• Handling clusters of different shapes
• Providing cluster membership probabilities

Installation

npm install hdbscan-ts

Usage

import { HDBSCAN } from "hdbscan-ts";
const data = [
  [1.1, 2.1],
  [2.1, 1.1],
  [1.1, 1.1],
  [0.1, 1.1],
  [10.1, 11.1],
  [11.1, 10.1],
  [10.1, 10.1]
];
const hdbscan = new HDBSCAN({
  minClusterSize: 2
});
const labels = hdbscan.fit(data);
console.log(labels);
// [0, 0, 0, 0, 1, 1, 1]

API

HDBSCAN

Constructor Options

• minClusterSize (default: 5): Minimum size of clusters
• minSamples (default: 5): Minimum number of samples in neighborhood
• debugMode (default: false): Enable debug logging

Methods

• fit(data: number[][]): HDBSCAN
• labels_: number[]
• probabilities_: number[]

License

MIT