适用数据类型：标称型

使用Python进行文本分类：

从文本中构建词向量：

  1. def loadDataSet():
  2.     postingList = [['my', 'dog', 'has', 'flea', 'problems', 'help', 'please'],
  3.                    ['maybe', 'not', 'take', 'him', 'to', 'dog', 'park', 'stupid'],
  4.                    ['my', 'dalmation', 'is', 'so', 'cute', 'I', 'love', 'him'],
  5.                    ['stop', 'posting', 'stupid', 'worthless', 'garbage'],
  6.                    ['mr', 'licks', 'ate', 'my', 'steak', 'how', 'to', 'stop', 'him'],
  7.                    ['quit', 'buying', 'worthless', 'dog', 'food', 'stupid']]
  8.     classVec = [0, 1, 0, 1, 0, 1]    #1 is abusive, 0 not
  9.     return postingList, classVec
  11. #利用集合set创造不重复的词组列表：
  12. def createVocabList(dataSet):
  13.     vocabSet = set([])  #create empty set
  14.     for document in dataSet:
  15.         vocabSet = vocabSet | set(document) #union of the two sets
  16.     return list(vocabSet)
  18. def setOfWords2Vec(vocabList, inputSet):
  19.     returnVec = [0]*len(vocabList)
  20.     for word in inputSet:
  21.         if word in vocabList:
  22.             returnVec[vocabList.index(word)] = 1
  23.         else: print("the word: %s is not in my Vocabulary!" % word)
  24.     return returnVec

image.png

image.png

  1. import numpy as np
  3. def trainNB0(trainMatrix, trainCategory):
  4.     numTrainDocs = len(trainMatrix)
  5.     numWords = len(trainMatrix[0])
  6.     pAbusive = sum(trainCategory)/float(numTrainDocs)
  7.     p0Num = np.ones(numWords); p1Num = np.ones(numWords)      #change to np.ones()
  8.     p0Denom = 2.0; p1Denom = 2.0                        #change to 2.0
  9.     for i in range(numTrainDocs):
  10.         if trainCategory[i] == 1:
  11.             p1Num += trainMatrix[i]
  12.             p1Denom += sum(trainMatrix[i])
  13.         else:
  14.             p0Num += trainMatrix[i]
  15.             p0Denom += sum(trainMatrix[i])
  16.     p1Vect = np.log(p1Num/p1Denom)          #change to np.log()
  17.     p0Vect = np.log(p0Num/p0Denom)          #change to np.log()
  18.     return p0Vect, p1Vect, pAbusive


image.png

修改分类器：

下溢出问题：使用对乘积取自然对数的方法。

  1. def classifyNB(vec2Classify, p0Vec, p1Vec, pClass1):
  2.     p1 = sum(vec2Classify * p1Vec) + np.log(pClass1)    #element-wise mult
  3.     p0 = sum(vec2Classify * p0Vec) + np.log(1.0 - pClass1)
  4.     if p1 > p0:
  5.         return 1
  6.     else:
  7.         return 0
  10.     def testingNB():
  11.     listOPosts, listClasses = loadDataSet()
  12.     myVocabList = createVocabList(listOPosts)
  13.     trainMat = []
  14.     for postinDoc in listOPosts:
  15.         trainMat.append(setOfWords2Vec(myVocabList, postinDoc))
  16.     p0V, p1V, pAb = trainNB0(np.array(trainMat), np.array(listClasses))
  17.     testEntry = ['love', 'my', 'dalmation']
  18.     thisDoc = np.array(setOfWords2Vec(myVocabList, testEntry))
  19.     print(testEntry, 'classified as: ', classifyNB(thisDoc, p0V, p1V, pAb))
  20.     testEntry = ['stupid', 'garbage']
  21.     thisDoc = np.array(setOfWords2Vec(myVocabList, testEntry))
  22.     print(testEntry, 'classified as: ', classifyNB(thisDoc, p0V, p1V, pAb))

image.png

  1. def bagOfWords2VecMN(vocabList, inputSet):
  2.     returnVec = [0]*len(vocabList)
  3.     for word in inputSet:
  4.         if word in vocabList:
  5.             returnVec[vocabList.index(word)] += 1
  6.     return returnVec

（词集模型：每个词出现或不出现；词袋模型，每个词出现的次数多少）

示例：使用朴素贝叶斯过滤垃圾邮件

image.png

问题：为什么没办法实现字符的分隔

可以换一种方法
image.png
https://www.cnblogs.com/xiaokuangnvhai/p/11213308.html
https://www.cnblogs.com/xp1315458571/p/13720333.html

  1. def textParse(bigString):    #input is big string, #output is word list
  2.     import re
  3.     listOfTokens = re.split(r'\W+', bigString)
  4.     return [tok.lower() for tok in listOfTokens if len(tok) > 2]
  6. def spamTest():
  7.     docList = []; classList = []; fullText = []
  8.     for i in range(1, 26):
  9.         wordList = textParse(open('email/spam/%d.txt' % i, encoding="ISO-8859-1").read())
  10.         docList.append(wordList)
  11.         fullText.extend(wordList)
  12.         classList.append(1)
  13.         wordList = textParse(open('email/ham/%d.txt' % i, encoding="ISO-8859-1").read())
  14.         docList.append(wordList)
  15.         fullText.extend(wordList)
  16.         classList.append(0)
  17.     vocabList = createVocabList(docList)#create vocabulary
  18.     trainingSet = range(50); testSet = []           #create test set
  19.     for i in range(10):
  20.         randIndex = int(np.random.uniform(0, len(trainingSet)))
  21.         testSet.append(trainingSet[randIndex])
  22.         del(list(trainingSet)[randIndex])
  23.     trainMat = []; trainClasses = []
  24.     for docIndex in trainingSet:#train the classifier (get probs) trainNB0
  25.         trainMat.append(bagOfWords2VecMN(vocabList, docList[docIndex]))
  26.         trainClasses.append(classList[docIndex])
  27.     p0V, p1V, pSpam = trainNB0(np.array(trainMat), np.array(trainClasses))
  28.     errorCount = 0
  29.     for docIndex in testSet:        #classify the remaining items
  30.         wordVector = bagOfWords2VecMN(vocabList, docList[docIndex])
  31.         if classifyNB(np.array(wordVector), p0V, p1V, pSpam) != classList[docIndex]:
  32.             errorCount += 1
  33.             print("classification error", docList[docIndex])
  34.     print('the error rate is: ', float(errorCount)/len(testSet))
  35.     #return vocabList, fullText

示例：从个人广告中获取区域倾向