kmeans

k-means clustering algorithm implementation in Go.

Usage

Only has two functions, Partition and PartitionWithCentroids.

Use PartitionWithCentroids if you want to partition observations with your initial centroids. This can be used if you want to calculate initial centroids with any custom logic.

Use Partition if you don't care about the creation of initial centroids and they will be randomly genarated based on observations.

Example

package main

import (
	"fmt"

	"github.com/mojixcoder/kmeans"
)

// Implements kmeans.Observation
type Location struct {
	Lng, Lat float64
}

func (l Location) X() float64 {
	return l.Lng

}

func (l Location) Y() float64 {
	return l.Lat
}

func main() {
	observations := []kmeans.Observation{
		Location{Lng: 1, Lat: 1},
		Location{Lng: 1, Lat: 2},
		Location{Lng: 2, Lat: 1},
		Location{Lng: 2, Lat: 2},

		Location{Lng: 14, Lat: 14},
		Location{Lng: 15, Lat: 14},
		Location{Lng: 14, Lat: 15},
		Location{Lng: 15, Lat: 15},

		Location{Lng: -10, Lat: -10},
		Location{Lng: -11, Lat: -10},
		Location{Lng: -10, Lat: -11},
		Location{Lng: -11, Lat: -11},
	}

	maxIters := 100

	k := 3

	clusters, err := kmeans.Partition(observations, maxIters, k)
	if err != nil {
		panic(err)
	}

	for _, cluster := range clusters {
		fmt.Println(cluster)
	}
}