multilabelsvm

This library allows svm(support vector mechine) to support multiclasses in nodejs and browser.

installation

node

npm install multilabelsvm

for svm kernel option refer svmjs

initialize the classifier as the following

	var multilabel = require('multilabelsvm' );
	var actionClassifier = new multilabel.Classifier({kernel : 'linear'});

browser

You need to include svmjs for this.

// include the library
<script src="./svmjs/lib/svm.js"></script>
<script src="./lib/multilabelsvm.js"></script>
<script>

var actionClassifier = new Classifier({
										kernel : 'rbf',
										C : 1.0,
										feature:{
											ngrams:2,
											casesensitive:true}
										});


</script>

Usage

Example usages are given below. svm configuration you can use all the parameter specified by the svmjs

in features option it suports ngrams,casesensitive

var trainSet = [
{ input:'What is your name',output: "name" },
{ input:'how are you',output: "fine"},
{ input:'please tell your name please',output: "name" },
{ input:'your name please',output: "name" },
{ input:'what is your name',output: "name" },
{ input:'who am i',output: "listener" },
{ input:'who are you ',output: "name" },
{ input:'may i know your name',output: "name" },
{ input:'your name',output: "name" },
{ input:'where you coming from',output: "about" },
{ input:'how do you do',output: "fine" },
{ input:'how are you doing',output: "fine"},
{ input:'how are you',output: "fine"},
{ input:'how do you do',output: "fine"},
{ input:'how are you',output: "fine"},
{ input:'what do you do',output: "fine"},
{ input:'can you edit this',output: "edit"},


]

actionClassifier.trainBatch(trainSet);



console.log(actionClassifier.classify('who are you'))
console.log(actionClassifier.classify('how are you'))

//backing up
var json = actionClassifier.toJSON()
var newActionClassifier = new multilabel.Classifier();
console.log('----------New Classifier----');
//importing
newActionClassifier.fromJSON(json);
console.log(newActionClassifier.classify('how are you'));
console.log(newActionClassifier.classify('who are you'));