goClustering

README

goClustering

goClustering is an implementation of following clustering algorithm in Golang.

Ward's method

Ward's method is a distance criterion in hierarchical cluster algorithm. This method puts two groups into one to minimize the total within-cluster variance. For detail, see here.

How to Install

You can get goClustering by using go get:

go get github.com/cipepser/goClustering/...

To visualize the dendrogram, you also install gonum/plot.

go get gonum.org/v1/plot/...

How to Use

This hierarchical cluster algorithm have two parts:

1. Execute Ward's method
2. Visualize

In execute Ward's method step, your input data must be n * d matrix, where n is an observation and d is a dimension of feature vectors. As a result of this step, you can get the ward.Tree which is binary-tree of the result of hierarchical clustering with distance of each group. You can execute Ward's method like this:

T := ward.Ward(X)

Second stap is Visualize step. You can visualize the dendrogram of ward.Tree by gonum/plot like:

d, _ := plotter.NewDendrogram(T)

Where, T is a result of first step. And you create a plot to add above d.

p, err := plot.New()
p.Add(d)

You can also customize title and labels of x-axis and y-axis.

p.Title.Text = "Dendrogram"
p.X.Label.Text = "data"
p.Y.Label.Text = "distance"

Name the leaf nodes and rotate names.

p.NominalX("aaa", "bbb", "ccc", "ddd", "eee", "fff")
p.X.Tick.Label.Rotation = math.Pi / 3
p.X.Tick.Label.YAlign = draw.YCenter
p.X.Tick.Label.XAlign = draw.XRight

Example

package main

import (
	"math"

	"github.com/cipepser/goClustering/ward"
	"github.com/cipepser/goClustering/vis"

	"gonum.org/v1/plot"
	"gonum.org/v1/plot/vg"
	"gonum.org/v1/plot/vg/draw"
)

func main() {
	// input data set
	X := [][]float64{
		{0, 0},
		{2, 2},
		{1, 1},
		{2, -1.2},
		{3, 2.2},
		{3.5, 0.5},
	}

	// Ward's method
	T := ward.Ward(X)

	// draw the dendrogram
	d, err := plotter.NewDendrogram(T)
	if err != nil {
		panic(err)
	}

	p, err := plot.New()
	if err != nil {
		panic(err)
	}
	p.Add(d)

	p.Title.Text = "Dendrogram"
	p.X.Label.Text = "data"
	p.NominalX("aaa", "bbb", "ccc", "ddd", "eee", "fff")

	p.X.Tick.Label.Rotation = math.Pi / 3
	p.X.Tick.Label.YAlign = draw.YCenter
	p.X.Tick.Label.XAlign = draw.XRight

	p.Y.Label.Text = "distance"

	// save as a png file
	file := "img.png"
	if err = p.Save(10*vg.Inch, 6*vg.Inch, file); err != nil {
		panic(err)
	}
}

Result

References

• Ward's method