JS implements Bayesian classifier

This time I will bring you JS to implement the Bayesian classifier. What are the precautions for JS to implement the Bayesian classifier? The following is a practical case, let's take a look.

First put the code

function NB(data) {    this.fc = {}; //记录特征的数量 feature conut 例如 {a:{yes:5,no:2},b:{yes:1,no:6}}
    this.cc = {}; //记录分类的数量 category conut 例如 {yes:6,no:8} }
NB.prototype = {
    infc(w, cls) { //插入新特征值
        if (!this.fc[w]) this.fc[w] = {};        
        if (!this.fc[w][cls]) this.fc[w][cls] = 0;        
        this.fc[w][cls] += 1;
    },
    incc(cls) { //插入新分类
        if (!this.cc[cls]) this.cc[cls] = 0;        
        this.cc[cls] += 1;
    },
    allco() { //计算分类总数 all count
        var t = 0;        
        for (var k in this.cc) t += this.cc[k];        
        return t;
    },
    fprob(w, ct) { //特征标识概率   
        if (Object.keys(this.fc).indexOf(w) >= 0) {            
        if (Object.keys(this.fc[w]).indexOf(ct) < 0) {                
        this.fc[w][ct] = 0
            }            
            var c = parseFloat(this.fc[w][ct]);            
            return c / this.cc[ct];
        } else {            
        return 0.0;
        }
    },
    cprob(c) { //分类概率
        return parseFloat(this.cc[c] / this.allco());
    },
    train(data, cls) { //参数:学习的Array,标识类型(Yes|No)
        for (var w of data) this.infc(String(w), cls);        
        this.incc(cls);
    },
    test(data) {        var ccp = {}; //P(类别)
        var fccp = {}; //P(特征|类别)
        for (var k in this.cc) ccp[k] = this.cprob(k);        
        for (var i of data) {
            i = String(i);            
            if (!i) continue;            
            if (Object.keys(this.fc).indexOf(i)) {                
            for (var k in ccp) {                    
            if (!fccp[k]) fccp[k] = 1;
                    fccp[k] *= this.fprob(i, k); //P(特征1|类别1)*P(特征2|类别1)*P(特征3|类别1)...
                }
            }
        }        var tmpk = "";        
        for (var k in ccp) {
            ccp[k] = ccp[k] * fccp[k];            
            if (!tmpk) tmpk = k;            
            if (ccp[k] > ccp[tmpk]) tmpk = k;
        }        
        return tmpk;
    }
};

The prediction function requires the use of the Naive Bayes algorithm

First of all, let’s look at the Bayesian formula:

Maybe you don’t understand the formula or understand the formula but don’t know how to use the formula

Then let me briefly translate it:

P( Category |Feature) = P ( Feature | Category ) * P( Category)/ P(Feature)

In fact, it is also:

P(category|feature)=P(feature|category)*P(category)/p(feature)

So we only need to calculate the following data:

P(feature|category)

P(category)
p(feature)

Assume two categories, category 1 and category 2

, then the total number of categories is The sum of the number of occurrences of the two categories
plus the possibility that the features we input have multiple hypotheses is just 3 simple:

P((Feature 1, Feature 2, Feature 3)|Category 1) = P(feature 1|category 1)*P(feature 2|category 1)*P(feature 3|category 1)

P(category 1)=number of categories 1/(total number of categories)
P( Feature 1, Feature 2, Feature 3) = P (Feature 1) * P ( Feature 2) * P ( Feature 3)

Because according to the formula we know:

P( Category 1|Feature)=P(Feature|Category 1)*P(Category 1)/p(Feature)

P(Category 2|Feature)=P(Feature|Category 2)*P(Category 2)/p (Feature)

It happens that p(Feature) is the denominator, so if you compare the probability of P(Category 1|Feature) and P(Category 2|Feature)

Just compare P(Feature|Category 1 )*P(Category 1) and P(Feature|Category 2)*P(Category 2) will be fine Other related articles!

Related reading:

How to use canvas to make a useful graffiti drawing board

How to use s- Merge cells in xlsx

The above is the detailed content of JS implements Bayesian classifier. For more information, please follow other related articles on the PHP Chinese website!