ac-learn/dist/evaluate.js

ML platform for All Contributors

"use strict";

// const ck = require('chalk')
const {
  PrecisionRecall
} = require('limdu').utils;

const ConfusionMatrix = require('./confusionMatrix');

const evalu = ({
  classifier,
  test,
  classes
  /* , log = true */

}) => {
  const curStats = new PrecisionRecall();
  const stats = {
    total: 0,
    TP: 0,
    TN: 0,
    FP: 0,
    FN: 0,
    confusionMatrix2D: []
  };
  const actual = []; //t.map(t => t.output)

  const predicted = [];
  test.forEach(t => {
    const predictedClasses = classifier.classify(t.input);
    /* const expl =  */

    curStats.addCases([t.output], predictedClasses); // console.log(`explanations (on ${t.input}->${t.output})= ${expl.join('\t')}`)

    let tn = true; // console.log(`expected "${ck.green.bold(t.output)}" on "${ck.cyan(t.input)}" and got: "${ck.yellow(predictedClasses.join('/'))}"`)
    // predictedClasses.forEach(pc => {
    //   stats[pc === t.output ? 'TP' : 'FP']++
    //   tn = false
    // })

    if (predictedClasses[0] === t.output) {
      tn = false;
      stats.TP++;
    } else if (!predictedClasses.includes(t.output)) {
      tn = false;
      stats.FN++;
    } else if (predictedClasses.length) {
      tn = false;
      stats.FP++;
    }

    if (tn) stats.TN++;
    stats.total += Math.max(predictedClasses.length, 1);

    if (predictedClasses.length) {
      /* predictedClasses.forEach(pc => {
        actual.push(t.output)
        predicted.push(pc)
      }) */
      actual.push(t.output);
      predicted.push(predictedClasses[0]);
    } else {
      actual.push(t.output);
      predicted.push('null');
    }
  });
  curStats.calculateStats();
  stats.confusionMatrix2D = [[stats.TP, stats.FP], [stats.FN, stats.TN]];
  stats.confusionMatrix = ConfusionMatrix.fromData(actual, predicted, classes);
  return {
    curStats,
    stats
  };
};

module.exports = evalu;