机器学习实战-朴素贝叶斯

数据集参考自https://blog.csdn.net/c406495762/article/details/77341116

朴素贝叶斯：首先，何为朴素？朴素要求的是条件特征之间相互独立。我们都知道大名鼎鼎的贝叶斯公式，其实朴素贝叶斯的思想很简单。就是通过计算属于某一类别的后验概率，然后比较大小，哪一类的概率大，那么就将它划分为哪一类。。。

由上述公式可知，我们先验概率P(A)很容易求得，我们重点在与求条件概率。由于条件特征之间相互独立，于是可以拆分成累乘的形式。在代码实现中，我们一般不会去求P(B)，因为分母都是一样的，我们只关注类别概率的大小！

将代码记录如下，方便以后查看。

import numpy as np
import pandas as pd
def getDataSet():
    postingList=[[‘my‘, ‘dog‘, ‘has‘, ‘flea‘, ‘problems‘, ‘help‘, ‘please‘],
                 [‘maybe‘, ‘not‘, ‘take‘, ‘him‘, ‘to‘, ‘dog‘, ‘park‘, ‘stupid‘],
                 [‘my‘, ‘dalmation‘, ‘is‘, ‘so‘, ‘cute‘, ‘I‘, ‘love‘, ‘him‘],
                 [‘stop‘, ‘posting‘, ‘stupid‘, ‘worthless‘, ‘garbage‘],
                 [‘mr‘, ‘licks‘, ‘ate‘, ‘my‘, ‘steak‘, ‘how‘, ‘to‘, ‘stop‘, ‘him‘],
                 [‘quit‘, ‘buying‘, ‘worthless‘, ‘dog‘, ‘food‘, ‘stupid‘]]
    labels = [0,1,0,1,0,1]
    return postingList,labels

#创建词汇表
def createVocabList(dataSet):
    vocabSet = set([])
    for data in dataSet:
        vocabSet = vocabSet | set(data)
    return list(vocabSet)

#向量化
def vectorize(vocabSet,dataSet):
    vocab = [0] * len(vocabSet)
    for data in dataSet:
        vocab[vocabSet.index(data)] = 1
    return vocab

#朴素贝叶斯建模
def trainN(X_train,y_train):
    num = len(X_train)   #有多少记录
    numvocab = len(X_train[0]) #词向量的大小
    p0Num = np.ones(numvocab) #统计非侮辱类的相关单词频数 加入了拉普拉斯平滑
    p1Num = np.ones(numvocab) #统计侮辱类的相关单词频数
    p0Sum = 2
    p1Sum = 2
    pA = sum(y_train) / num                   #先验概率
    for i in range(num):
        if y_train[i]==0:   #统计属于非侮辱类的条件概率所需的数据
            p0Sum += sum(X_train[i])
            p0Num += X_train[i]
        else:               #统计属于侮辱类的条件概率所需的数据
            p1Sum += sum(X_train[i])
            p1Num += X_train[i]

    # 为了防止下溢出，计算条件概率的对数
    p0 = np.log(p0Num / p0Sum)      #频数除以总数 得到概率
    p1 = np.log(p1Num / p1Sum)
    return p0,p1,pA

#分类
def classify(testMat,p0,p1,pA):
    p0Score = sum(testMat * p0) + np.log(pA)
    p1Score = sum(testMat * p1) + np.log(1-pA)
    if p0Score > p1Score:
        return 0
    else:
        return 1

if __name__==‘__main__‘:
    dataSet,label = getDataSet()
    vocabSet = createVocabList(dataSet)
    trainMat = []
    for elem in dataSet:
        trainMat.append(vectorize(vocabSet,elem))
    # print(trainMat)
    p0,p1,pA = trainN(trainMat,label)
    test1= [‘love‘, ‘my‘, ‘dalmation‘]
    test2= [‘stupid‘, ‘garbage‘]
    test1_vocab = np.array(vectorize(vocabSet,test1))
    test2_vocab = np.array(vectorize(vocabSet,test2))
    result1 = classify(test1_vocab,p0,p1,pA)
    result2 = classify(test2_vocab,p0,p1,pA)
    if result1==1:
        print(test1,"属于：侮辱类")
    else:
        print(test1, "属于：非侮辱类")
    print("=======================================")
    if result2==1:
        print(test2,"属于：侮辱类")
    else:
        print(test2, "属于：非侮辱类")

结果如下：[‘love‘, ‘my‘, ‘dalmation‘] 属于：非侮辱类
=======================================
[‘stupid‘, ‘garbage‘] 属于：侮辱类

原文地址：https://www.cnblogs.com/logo-88/p/10146509.html