kyrema

A simple k-means clustering implement.

Usage

import { kyrema } from 'kyrema'

const dataset = [0, 1, 2, 5, 6, 7, 8, 9]

const centroids = kyrema(
  2,
  dataset,
  (d, c) => Math.abs(d - c), // calculate the distance
  (datas) => datas.reduce((p, c) => p + c, 0) / datas.length, // calculate the mean
  (c1, c2) => c1 === c2, // check if equal
)
console.log(centroids) // The value of centroids: 1, 7.

API

• kyrema(k, data, distanceCalculator, averageCalculator, centroidEqualator, maxTry)

  • Parameters:
    • k - number
      • The cluster count.
    • data - T[]
      • The dataset, where T is a generic type.
    • distanceCalculator - (data: T, centroid: T) => number
      • Determine how to calculate the distance between a data and the centroid.
    • averageCalculator - (datas: T[]) => T
      • Determine how to calculate the mean value of a cluster.
    • centroidEqualator - (c1: T, c2: T) => boolean
      • Determine how to check if the two datas are equal to each other.
    • maxTry - number
      • [Optional] The maximum repeat times of the algorithm. Default: 50.
  • Return:
    • Centroid[]
      • An array with k centroids.

• kyremaWithCentroids(k, initialCentroids, data, distanceCalculator, averageCalculator, centroidEqualator, maxTry)

  • Parameters:
    • initialCentroids - T[]
      • Initial centroids, an array with k elements.
    • Note: Other parameters are the same as the above function: kyrema
  • Return:
    • Centroid[]
      • An array with k centroids.

• Centroid

  • The returned object of the algorithm, with the following properties:
    • value - T
      • The value of the cluster mean. It's of type T.
    • count - number
      • The amount of items in the cluster.
    • indexes - number[]
      • The index of each item in the original dataset.

License

MIT