朴素贝叶斯

最新推荐文章于 2023-01-27 22:00:47 发布

酸乳酸乳

最新推荐文章于 2023-01-27 22:00:47 发布

阅读量184

点赞数

CC 4.0 BY-SA版权

分类专栏：机器学习

本文链接：https://blog.youkuaiyun.com/qq_41359265/article/details/91125476

机器学习专栏收录该内容

5 篇文章

订阅专栏

本文深入探讨了贝叶斯分类器中的朴素贝叶斯算法，这是一种简单且高效的分类方法，尤其适用于文本分类等场景。文章通过实例展示了算法的工作原理，包括数据集的加载、词汇表的创建、词向量的转换以及训练过程，最后通过测试数据验证了分类器的准确性。

摘要生成于 C知道，由 DeepSeek-R1 满血版支持，前往体验 >

贝叶斯分类是一类分类算法的总称，这类算法均以贝叶斯定理为基础，故统称为贝叶斯分类。而朴素贝叶斯分类器是贝叶斯分类器中最简单常用的一种方法。并且朴素贝叶斯算法仍然是最流行的十大挖掘算法之一。该算法简单易懂，学习效率高，在某些分类问题上能够和神经网络和决策树相媲美。

import numpy as np

def load_data_set():
    posting_list = [
        ['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']]
    class_vec = [0, 1, 0, 1, 0, 1]  # 1 is 侮辱性的文字, 0 is not
    return posting_list, class_vec

def createVocabList(postingList):
    result = set({})
    for posting in postingList:
        result = result|set(posting)
    return list(result)

def setOfWord2Vec(vocablist,inputset):
    result = [0]*len(vocablist)
    for input in inputset:
        if input in vocablist:
            result[vocablist.index(input)] = 1
        else:
            print(input,"is not in vocablist")
    return result

#朴素贝叶斯词袋模型 词的频率可能携带某些信息
def bagOfWord2Vec(vocabList,inpuSet):
    result = [0] * len(vocabList)
    for input in inpuSet:
        if input in vocabList:
            result[vocabList.index(input)] += 1
        else:
            print(input,"is not in vocablist")
    return  result

def trainNB(trainMatrix,classVec):
    numPost = len(trainMatrix)
    numWord = len(trainMatrix[0])
    pc = np.sum(classVec) / np.float(numPost)
    num0Vec = np.ones(numWord)
    num1Vec = np.ones(numWord)
    num0 = 2.0
    num1 = 2.0
    for i in range(numPost):
        if classVec[i] == 1:
            num1Vec += trainMatrix[i]
            num1 += np.sum(trainMatrix[i])
        else:
            num0Vec += trainMatrix[i]
            num0 += np.sum(trainMatrix[i])
    p1 = np.log(num1Vec / num1)
    p0 = np.log(num0Vec / num0)
    return p0,p1,pc

def classifier(vec2classifier,p0Vec,p1Vec,pc):
    p0 = np.sum(vec2classifier * p0Vec) + np.log(1.0 - pc)
    p1 = np.sum(vec2classifier * p1Vec) + np.log(pc)
    print(p0,p1)
    if p0 < p1:
        return 1
    else:
        return 0


def testingNB():
    listPosts,listClasses = load_data_set()
    myVocab = createVocabList(listPosts)
    trainMat = []
    for i in listPosts:
        trainMat.append(setOfWord2Vec(myVocab,i))
    p0,p1,pc = trainNB(np.array(trainMat),np.array(listClasses))
    p0 = np.array(p0)
    p1 = np.array(p1)
    test0 = ['love', 'my', 'dalmation']
    test1 = ['stupid', 'garbage']
    test0vect = np.array(setOfWord2Vec(myVocab, test0))
    test1vect = np.array(setOfWord2Vec(myVocab, test1))
    print("test0 ：", classifier(test0vect, p0, p1, pc))
    print("test1 :", classifier(test1vect, p0, p1, pc))



if __name__ == "__main__":
    #testingNB()
    pass