【JavaScript】Naive Bayes

相似文件 换一批

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Naive Bayes

Naive bayes classifier.

Methodsnew NaiveBayes()

Constructor that takes no arguments.

Example

var nb = new NaiveBayes();
train(trainingSet, predictions)

Train the Naive Bayes model to the given training set and predictions

Arguments

• trainingSet - A matrix of the training set.
• trainingLabels - An array of value for each case in the training set.
  

本帖隐藏的内容

naive-bayes-master.zip (8.01 KB)

2017-4-16 04:11:19 上传

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Javascript script naive Bayes scrip training matrix

1. "use strict";
   
2. 
   
3. var Matrix = require('ml-matrix');
   
4. var Utils = require('./utils');
   
5. 
   
6. class MultinomialNB {
   
7.     constructor(model) {
   
8.     }
   
9. 
   
10.     train(trainingSet, trainingLabels) {
    
11.         trainingSet = Matrix.checkMatrix(trainingSet);
    
12.         var separateClasses = Utils.separateClasses(trainingSet, trainingLabels);
    
13.         this.priorProbability = new Matrix(separateClasses.length, 1);
    
14. 
    
15.         for(var i = 0; i < separateClasses.length; ++i) {
    
16.             this.priorProbability[i][0] = Math.log(separateClasses[i].length / trainingSet.rows);
    
17.         }
    
18. 
    
19.         var features = trainingSet.columns;
    
20.         this.conditionalProbability = new Matrix(separateClasses.length, features);
    
21.         for(i = 0; i < separateClasses.length; ++i) {
    
22.             var classValues = Matrix.checkMatrix(separateClasses[i]);
    
23.             var total = classValues.sum();
    
24.             var divisor = total + features;
    
25.             this.conditionalProbability.setRow(i, classValues.sum('column').add(1).div(divisor).apply(matrixLog));
    
26.         }
    
27.     }
    
28. 
    
29.     predict(dataset) {
    
30.         dataset = Matrix.checkMatrix(dataset);
    
31.         var predictions = new Array(dataset.rows);
    
32.         for(var i = 0; i < dataset.rows; ++i) {
    
33.             var currentElement = dataset.getRowVector(i);
    
34.             predictions[i] = this.conditionalProbability.clone().mulRowVector(currentElement).sum('row')
    
35.                              .add(this.priorProbability).maxIndex()[0];
    
36.         }
    
37.     }
    
38. 
    
39.     toJSON() {
    
40. 
    
41.     }
    
42. 
    
43.     load() {
    
44. 
    
45.     }
    
46. }
    
47. 
    
48. function matrixLog(i, j) {
    
49.     this[i][j] = Math.log(this[i][j]);
    
50. }
    
51. 
    
52. module.exports = MultinomialNB;

复制代码

1. 'use strict';
   
2. 
   
3. var Matrix = require('ml-matrix');
   
4. var Stat = require('ml-stat');
   
5. var Utils = require('./utils');
   
6. 
   
7. module.exports = NaiveBayes;
   
8. //module.exports.separateClasses = separateClasses;
   
9. 
   
10. /**
    
11. * Constructor for the Naive Bayes classifier, the parameters here is just for loading purposes.
    
12. *
    
13. * @param reload
    
14. * @param model
    
15. * @constructor
    
16. */
    
17. function NaiveBayes(reload, model) {
    
18.     if(reload) {
    
19.         this.means = model.means;
    
20.         this.calculateProbabilities = model.calculateProbabilities;
    
21.     }
    
22. }
    
23. 
    
24. /**
    
25. * Function that trains the classifier with a matrix that represents the training set and an array that
    
26. * represents the label of each row in the training set. the labels must be numbers between 0 to n-1 where
    
27. * n represents the number of classes.
    
28. *
    
29. * WARNING: in the case that one class, all the cases in one or more features have the same value, the
    
30. * Naive Bayes classifier will not work well.
    
31. * @param trainingSet
    
32. * @param trainingLabels
    
33. */
    
34. NaiveBayes.prototype.train = function (trainingSet, trainingLabels) {
    
35.     var C1 = Math.sqrt(2*Math.PI); // constant to precalculate the squared root
    
36.     if(!Matrix.isMatrix(trainingSet)) trainingSet = new Matrix(trainingSet);
    
37.     else trainingSet = trainingSet.clone();
    
38. 
    
39.     if(trainingSet.rows !== trainingLabels.length)
    
40.         throw new RangeError("the size of the training set and the training labels must be the same.");
    
41. 
    
42.     var separatedClasses = Utils.separateClasses(trainingSet, trainingLabels);
    
43.     var calculateProbabilities = new Array(separatedClasses.length);
    
44.     this.means = new Array(separatedClasses.length);
    
45.     for(var i = 0; i < separatedClasses.length; ++i) {
    
46.         var means = Stat.matrix.mean(separatedClasses[i]);
    
47.         var std = Stat.matrix.standardDeviation(separatedClasses[i], means);
    
48. 
    
49.         var logPriorProbability = Math.log(separatedClasses[i].rows / trainingSet.rows);
    
50.         calculateProbabilities[i] = new Array(means.length + 1);
    
51. 
    
52.         calculateProbabilities[i][0] = logPriorProbability;
    
53.         for(var j = 1; j < means.length + 1; ++j) {
    
54.             var currentStd = std[j - 1];
    
55.             calculateProbabilities[i][j] = [(1 / (C1 * currentStd)), -2*currentStd*currentStd];
    
56.         }
    
57. 
    
58.         this.means[i] = means;
    
59.     }
    
60. 
    
61.     this.calculateProbabilities = calculateProbabilities;
    
62. };
    
63. 
    
64. /**
    
65. * function that predicts each row of the dataset (must be a matrix).
    
66. *
    
67. * @param dataset
    
68. * @returns {Array}
    
69. */
    
70. NaiveBayes.prototype.predict = function (dataset) {
    
71.     if(dataset[0].length === this.calculateProbabilities[0].length)
    
72.         throw new RangeError('the dataset must have the same features as the training set');
    
73. 
    
74.     var predictions = new Array(dataset.length);
    
75. 
    
76.     for(var i = 0; i < predictions.length; ++i) {
    
77.         predictions[i] = getCurrentClass(dataset[i], this.means, this.calculateProbabilities);
    
78.     }
    
79. 
    
80.     return predictions;
    
81. };
    
82. 
    
83. /**
    
84. * Function the retrieves a prediction with one case.
    
85. *
    
86. * @param currentCase
    
87. * @param mean - Precalculated means of each class trained
    
88. * @param classes - Precalculated value of each class (Prior probability and probability function of each feature)
    
89. * @returns {number}
    
90. */
    
91. function getCurrentClass(currentCase, mean, classes) {
    
92.     var maxProbability = 0;
    
93.     var predictedClass = -1;
    
94. 
    
95.     // going through all precalculated values for the classes
    
96.     for(var i = 0; i < classes.length; ++i) {
    
97.         var currentProbability = classes[i][0]; // initialize with the prior probability
    
98.         for(var j = 1; j < mean[0].length + 1; ++j) {
    
99.             currentProbability += calculateLogProbability(currentCase[j - 1], mean[i][j - 1], classes[i][j][0], classes[i][j][1]);
    
100.         }
     
101. 
     
102.         currentProbability = Math.exp(currentProbability);
     
103.         if(currentProbability > maxProbability) {
     
104.             maxProbability = currentProbability;
     
105.             predictedClass = i;
     
106.         }
     
107.     }
     
108. 
     
109.     return predictedClass;
     
110. }
     
111. 
     
112. /**
     
113. * Function that export the NaiveBayes model.
     
114. * @returns {{modelName: string, means: *, calculateProbabilities: *}}
     
115. */
     
116. NaiveBayes.prototype.export = function () {
     
117.     return {
     
118.         modelName: "NaiveBayes",
     
119.         means: this.means,
     
120.         calculateProbabilities: this.calculateProbabilities
     
121.     };
     
122. };
     
123. 
     
124. /**
     
125. * Function that create a Naive Bayes classifier with the given model.
     
126. * @param model
     
127. * @returns {NaiveBayes}
     
128. */
     
129. NaiveBayes.load = function (model) {
     
130.     if(model.modelName !== 'NaiveBayes')
     
131.         throw new RangeError("The given model is invalid!");
     
132. 
     
133.     return new NaiveBayes(true, model);
     
134. };
     
135. 
     
136. /**
     
137. * function that retrieves the probability of the feature given the class.
     
138. * @param value - value of the feature.
     
139. * @param mean - mean of the feature for the given class.
     
140. * @param C1 - precalculated value of (1 / (sqrt(2*pi) * std)).
     
141. * @param C2 - precalculated value of (2 * std^2) for the denominator of the exponential.
     
142. * @returns {number}
     
143. */
     
144. function calculateLogProbability(value, mean, C1, C2) {
     
145.     var value = value - mean;
     
146.     return Math.log(C1 * Math.exp((value * value) / C2))
     
147. }

复制代码