data-clustering

clustering

A set of clustering tool for javascript

Installation

npm

npm install data-clustering

bower

bower install data-clustering

Usage

if using npm:

import * as cl from 'data-clustering';

if using bower

<script src="bower_components/clustering/dist/clustering.js"></script>

What's in it

Clustering Tools

Hierachical Cluster

Example:

	var hc_cluster = cl.HierachicalCluster()
	.data([{
		name : '1',
		value : {
			point : [0, 1]
		}
	},
	{
		name : '2',
		value: {
			point : [0, 2]
		}
	},
	{
		name : '3',
		value: {
			point : [0, 3]
		}
	},
	{
		name : '4',
		value : {
			point : [0, 3]
		}
	},
	{
		name : '5',
		value : {
			point : [0, -1]
		}
	},
	{
		name : '6',
		value : {
			point : [0, 10]
		}
	},
	{
		name : '7',
		value : {
			point : [0, 11]
		}
	}])
	.dist_metric(cl.euclidean_distance)
	.dist_fun('max')
	.save_history(true)
	.init()
	.cluster();

	console.log('root', hc_cluster.root());

Kmeans Cluster

Example:

	var km_cluster = cl.KMean()
	.data([{
		name : '1',
		value : {
			point : [1, 1]
		}
	},
	{
		name : '2',
		value: {
			point : [1.5, 2.0]
		}
	},
	{
		name : '3',
		value: {
			point : [3, 4]
		}
	},
	{
		name : '4',
		value : {
			point : [5, 7]
		}
	},
	{
		name : '5',
		value : {
			point : [3.5, 5]
		}
	},
	{
		name : '6',
		value : {
			point : [4.5, 5]
		}
	},
	{
		name : '7',
		value : {
			point : [3.5, 4.5]
		}
	}
	])
	.clusters([
		{
			'name' : 'C1',
			'value' : {
				'centroid' : [1, 1]
			}
		},
		{
			'name' : 'C2',
			'value' : {
				'centroid' : [5, 7]
			}
		}
	])
	.evaluate_sse(true)
	.save_history(true)
	.stopThreshold(0)
	.accessor(function(d){return d.value.point;})
	.centroid_fun('mean')
	.numIteration(4)
	.dist_metric(cl.euclidean_distance)
	.cluster();

Clustering Evaluation

Example:

	//create some points
	var points = cl.array2points([
			[0.4, 0.53],
			[0.22, 0.38],
			[0.35, 0.32],
			[0.26, 0.19],
			[0.08, 0.41],
			[0.45, 0.30]
		]);

	//perform hierachical clustering
	var cluster = cl.HierachicalCluster()
	.data(points)
	.dist_metric(cl.euclidean_distance)
	.dist_fun('centroid')
	.save_history(true)
	.init()
	.cluster();

	//Cut the hierachical clustering to 3 clusters
	var clustering = cluster.cut_opt('K').cut(3);

	//creating the clustering evaluation object
	var cev = cl.ClusterEvaluation().data(clustering);

	var wss = cev.WSS();
	var bss = cev.BSS();
	var tss = cev.TSS();
	var silhouette = cev
	.silhouette_dist_metric(cl.euclidean_distance)
	.silhouette_coefficient();

Sparse Vector

Example:

	var v1 = cl.SparseVector([0, 1, 5, 6, 10], [1, 1, 1, 1, 1]);
	var v2 = cl.SparseVector([0, 5, 10, 11], [1, 1, 1, 1]);
	var v3 = cl.SparseVector([1], [1]);
	var d = v1.dotp(v2);
	var s = v1.sum(v2);
	console.log('d', d);
	console.log('L2', v1.L2norm());
	console.log('s', s);
	var ind1 = v1.locationAtIndex(5);
	console.log('index1', ind1);
	var ind3 = v3.locationAtIndex(0);
	console.log('index3', ind3);
	console.log('v1', v1.toDenseVector());
	console.log('v2', v2.toDenseVector());
	console.log('v3', v3.toDenseVector());
	console.log('s', s.toDenseVector());

	v1.setValue(1, 2);
	v1.setValue(4, 1);
	v1.setValue(20, 1);
	console.log('v1', v1.toDenseVector());

Dijkstra's shortest path algorithm

Example:

	var nodes = [
	{
		id : 0,
		name : 0
	},
	{
		id: 1,
		name : 1
	},
	{
		id : 2,
		name : 2
	},
	{
		id : 3,
		name : 3
	},
	{
		id : 4,
		name : 4
	},
	{
		id : 5,
		name : 5
	},
	{
		id : 6,
		name : 6
	}
	];

	var edges = [
	{
		source : nodes[0],
		target : nodes[1],
		value : 2
	},
	{
		source : nodes[0],
		target : nodes[2],
		value : 9
	},
	{
		source : nodes[1],
		target : nodes[2],
		value : 4
	},
	{
		source : nodes[1],
		target : nodes[3],
		value : 2
	},
	{
		source : nodes[2],
		target : nodes[3],
		value : 1
	},
	{
		source : nodes[2],
		target : nodes[5],
		value : 3
	},
	{
		source : nodes[2],
		target : nodes[6],
		value : 11
	},
	{
		source : nodes[3],
		target : nodes[4],
		value : 1
	},
	{
		source : nodes[4],
		target : nodes[6],
		value : 7
	},
	{
		source : nodes[5],
		target : nodes[6],
		value : 7
	}
	];

	var G = cl.Graph().nodes(nodes).edges(edges).create();

	var dk = cl.ShortestPathDijkstra()
	.direction('out')
	.init_metric(function(){return 0;})
	.init_source_metric(function(){return Infinity;})
	.comparator(function(a, b){
		return b - a;
	})
	// .source(G.nodes()[0])
	.graph(G);

	var paths = dk();
	var path;
	var i;
	for(i = 0; i < paths.length; i++){
		path = paths[i];
		console.log(i, path.map(function(d){
			return d.id;
		}));
	}

Girvan Newman Network Clustering Althorithm

	var G = cl.Graph().nodes(nodes).edges(edges).create();
	var ge = cl.GirvanNewman().graph(G);
	var tree = ge();
	console.log('tree', tree);