exports.buildSVM = function(trainingSet, callback){
console.log('Building SVM');
// Get the parameters correct
// using cross-fold classification
var svm = new nodesvm.CSVC({
C: [0.03125, 0.125, 0.5, 2, 8],
// kernels parameters
kernelType: 'RBF',
gamma: [0.03125, 0.125, 0.5, 2, 8],
// training options
nFold: 5,
normalize: true,
reduce: false,
probability : true
});
svm.train(trainingSet).progress(function(rate){
//console.log('training progress: %d%', Math.round(rate*100));
})
.spread(function(trainedModel, trainingReport){
//console.log('SVM trained. \nReport:\n%s', so(trainingReport));
return callback(JSON.stringify(trainedModel));
});
};
'use strict';
const fs = require( 'fs' );
const svm = require( 'node-svm' );
const _ = require( 'lodash' );
const util = require( 'util' );
var data = require( '../model/data' );
var dataset = data.features.map( (e, i) => {
var o = [];
o[0] = e;
o[1] = data.sentiments[i];
return o;
});
var clf = new svm.CSVC({
kernelType: 'linear',
probability: true,
c:[0.005, 0.01,0.125,0.5,1,2]
});
clf.train(dataset)
.progress( function( rate ) {
console.log( rate );
})
.spread( (trainedModel, trainingReport) => {
console.log(trainingReport);
fs.writeFileSync( __dirname + '/../model/model.json', JSON.stringify(trainedModel) );
});