sketch.js

// Daniel Shiffman
// http://codingtra.in
// http://patreon.com/codingtrain

// Machine Learning
// Nearest Neighbor Recommendations
// More: https://github.com/shiffman/NOC-S17-2-Intelligence-Learning/tree/master/week3-classification-regression

// Part 1: https://youtu.be/N8Fabn1om2k
// Part 2: https://youtu.be/Lo89NLmSgl0
// Part 3: https://youtu.be/aMtckmWAzDg

var data;
var users;

var resultP;
var resultDivs = [];

function preload() {
  data = loadJSON('movies.json');
}

function setup() {
  noCanvas();
  users = {};

  var dropdowns = [];

  var titles = data.titles;
  for (var i = 0; i < titles.length; i++) {
    var div = createDiv(titles[i]);
    var dropdown = createSelect('');
    dropdown.title = titles[i];
    dropdown.option('not seen');
    dropdown.parent(div);
    dropdowns.push(dropdown);
    for (var star = 1; star < 6; star++) {
      dropdown.option(star);
    }
  }

  var button = createButton('submit');
  button.mousePressed(predictRatings);
  resultP = createP('');

  function predictRatings() {
    var newUser = {};
    for (var i = 0; i < dropdowns.length; i++) {
      var title = dropdowns[i].title;
      var rating = dropdowns[i].value();
      if (rating == 'not seen') {
        rating = null;
      }
      newUser[title] = rating;
    }

    findNearestNeighbors(newUser);
  }

  function findNearestNeighbors(user) {
    for (var i = 0; i < resultDivs.length; i++) {
      resultDivs[i].remove();
    }
    resultDivs = [];

    var similarityScores = {};
    for (var i = 0; i < data.users.length; i++) {
      var other = data.users[i];
      var similarity = euclideanDistance(user, other);
      similarityScores[other.name] = similarity;
    }

    data.users.sort(compareSimilarity);

    function compareSimilarity(a, b) {
      var score1 = similarityScores[a.name];
      var score2 = similarityScores[b.name];
      return score2 - score1;
    }

    for (var i = 0; i < data.titles.length; i++) {
      var title = data.titles[i];
      if (user[title] == null) {
        var k = 5;
        var weightedSum = 0;
        var similaritySum = 0;
        for (var j = 0; j < k; j++) {
          var name = data.users[j].name;
          var sim = similarityScores[name];
          var ratings = data.users[j];
          var rating = ratings[title];
          if (rating != null) {
            weightedSum += rating * sim;
            similaritySum += sim;
          }
        }

        var stars = nf(weightedSum / similaritySum, 1, 2);
        var div = createDiv(title + ': ' + stars);
        resultDivs.push(div);
        div.parent(resultP);
      }
    }

    // var k = 5;
    // for (var i = 0; i < k; i++) {
    //   var name = data.users[i].name;
    //   var score = nf(similarityScores[name], 1, 2)
    //   var div = createDiv(name + ': ' + score);
    //   resultDivs.push(div);
    //   div.parent(resultP);
    // }

    // console.log(similarityScores);
  }
}

function euclideanDistance(ratings1, ratings2) {
  var titles = data.titles;

  var sumSquares = 0;
  for (var i = 0; i < titles.length; i++) {
    var title = titles[i];
    var rating1 = ratings1[title];
    var rating2 = ratings2[title];
    if (rating1 != null && rating2 != null) {
      var diff = rating1 - rating2;
      sumSquares += diff * diff;
    }
  }
  var d = sqrt(sumSquares);

  var similarity = 1 / (1 + d);
  return similarity;
}