import numpy as np
from scipy import stats
import matplotlib.pyplot as plt
import math
import sys

#创建去重的词汇表
def create_vocabulary_list(dataset):
    # dataset 包含了多条留言的文本，每一条留言为一个列表
    vocabset = set([])
    for document in dataset:
        vocabset = vocabset | set(document)
    return list(vocabset)

def wordsToVector(vocabset, inputset):
    # vocabset 是词汇表， inputset 为要转化为向量的文本
    vector = [0]*len(vocabset)
    for word in inputset:
        if word in vocabset:
            vector[vocabset.index(word)] = 1
        else:
            print("The word: {} is not in my Vocabulary.".format(word))
    return vector

def traing_bayes(trainset, category):
    """sovle the conditional probility"""
    # trainCategory 是标记为负类和正类的向量
    num_train = len(trainset)    # 获取训练集的数量，其中训练集是由文本向量组成
    num_vocab = len(trainset[0]) # 由于训练集中的每个元素都是文本向量，即包含了整个词汇表

    pAbusive = sum(category) / float(num_train)# 在训练集中计算负类的概率，负类的记录数除以训练集总数

    #下面的这2个变量计算每个词汇在正负类下出现的次数，由于词汇表是训练集组成的不重复的词汇集，所以，每个词都来自词汇表，即至少会出现一次
    pnorm_vector = np.ones(num_vocab) # 每个词汇出现在正类中的概率组成的向量
    pabu_vector = np.ones(num_vocab) # 每个词汇出现在负类中的概率组成的向量

    pnorm_denom = 2.0 # 正类中词汇的数目
    pabu_denom = 2.0  # 负类中词汇的数目

    for i in range(num_train):
        if category[i] == 1: #负类
            pabu_vector += trainset[i]
            pabu_denom += sum(trainset[i])
        else: #正类
            pnorm_vector += trainset[i]
            pnorm_denom += sum(trainset[i])

    pnorm_vector = np.log(pnorm_vector / pnorm_denom)
    pabu_vector = np.log(pabu_vector / pabu_denom)

    return pnorm_vector, pabu_vector, pAbusive

def classify(testVector, pnorm_vector, pabu_vector, pabusive):
    # 先将测试文本转化为向量

    pnorm = sum(testVector*pnorm_vector) + np.log(1 - pabusive) # 计算测试文本为正类的概率
    pabu = sum(testVector*pabu_vector) + np.log(pabusive) # 计算测试文本为负类的概率

    if pabu > pnorm:
        return 1
    else:
        return 0


post_list = [
    ['my', 'dog', 'has', 'flea', 'problem', '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', 'ny', 'steak', 'how', 'to', 'stop', 'him'],
    ['quit', 'buying', 'worthless', 'dog', 'food', 'stupid']
    ] #训练集
category = [0, 1, 0, 1, 0, 1]

def test_bayes():

    vocab_list = create_vocabulary_list(post_list) #训练集去重复词汇表

    trainset = [] #训练集转向量
    for post in post_list:
        trainset.append(wordsToVector(vocab_list, post))

    pnorm_vector, pabu_vector, pAbusive = traing_bayes(trainset, category)#用训练集进行训练得出先验概率

    testEntry = ['love', 'my', 'dalmation']
    testVector = np.array(wordsToVector(vocab_list, testEntry))

    #判别
    print(testEntry, "classified as: {}".format(classify(testVector, pnorm_vector, pabu_vector, pAbusive)))

    testEntry = ['stupid', 'garbage']
    testVector = np.array(wordsToVector(vocab_list, testEntry))

    print(testEntry, "classified as: {}".format(classify(testVector, pnorm_vector, pabu_vector, pAbusive)))


test_bayes()

---

loading......