naiveBayes3.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 朴素贝叶斯之新浪新闻分类(Sklearn)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1 准备数据, 生成词汇表"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import jieba\n",
    "import random\n",
    "from sklearn.naive_bayes import MultinomialNB\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 将所有的单词生成词汇表\n",
    "# 读入所有文件以及label, 将读进来的数据分成训练集和测试集\n",
    "# 利用训练集中的数据生成词汇表, 并按单词出现的数量由高到底进行排序\n",
    "# 返回词汇表, 训练集, 测试集, 训练label, 测试label\n",
    "def TextProcessing(folder_path, test_size=0.2):\n",
    "    folder_list = os.listdir(folder_path)   # 存放所有数据的主文件夹路径\n",
    "    data_list = []\n",
    "    class_list = []\n",
    "    for folder in folder_list:             # 遍历主文件夹下的每一个文件夹, 也就是每一个类别\n",
    "        new_folder_path = os.path.join(folder_path, folder)\n",
    "        files = os.listdir(new_folder_path) # 当前文件夹下的所有文件\n",
    "        j = 1\n",
    "        for file in files:\n",
    "            if j > 100:\n",
    "                break\n",
    "            with open(os.path.join(new_folder_path, file), 'r', encoding='utf-8') as f:\n",
    "                raw = f.read()\n",
    "            word_cut = jieba.cut(raw, cut_all=False)     # 分词\n",
    "            word_list = list(word_cut)                   \n",
    "            data_list.append(word_list)\n",
    "            class_list.append(folder)\n",
    "            j+=1\n",
    "#     print(data_list)\n",
    "#     print(class_list)    \n",
    "    #      \n",
    "    data_class_list = list(zip(data_list, class_list))  # 将文件和对应的列表压缩在一起\n",
    "    random.shuffle(data_class_list)                     \n",
    "    index = int(len(data_class_list) * test_size) + 1   # 测试集的文件个数\n",
    "    train_list = data_class_list[index:]\n",
    "    test_list = data_class_list[:index]\n",
    "    train_data_list, train_class_list = zip(*train_list) # 解压缩, 分成文件和对应的label\n",
    "    test_data_list, test_class_list = zip(*test_list)\n",
    "    # 统计每一个单词的个数 生成键值对{'单词': num}\n",
    "    all_words_dict = {}\n",
    "    for word_list in train_data_list:\n",
    "        for word in word_list:\n",
    "            if word in all_words_dict.keys():\n",
    "                all_words_dict[word] += 1\n",
    "            else:\n",
    "                all_words_dict[word] = 1\n",
    "    all_words_tuple_list = sorted(all_words_dict.items(), key=lambda f:f[1], reverse=True) # 对值由高到底排序\n",
    "    all_word_list, all_word_nums = zip(*all_words_tuple_list)\n",
    "    all_word_list = list(all_word_list)\n",
    "    return all_word_list, train_data_list, test_data_list, train_class_list, test_class_list\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 制作一个set集合, 里面存储的是一些无用的高频词\n",
    "# 因为上一步生成的词汇表中有一些高频词是: 标点符号, '的 地 得', 以及数字, 我们要将这些去掉\n",
    "# 所以提前准备了一个文件, 里面包含要去掉的单词\n",
    "def makeWordSet(words_file):\n",
    "    word_set = set()\n",
    "    with open(words_file, 'r', encoding='utf-8') as f:\n",
    "        for line in f.readlines():\n",
    "            word = line.strip()     # 去掉这一行的空格\n",
    "            if len(word) > 0: \n",
    "                word_set.add(word)  # 放入set集合\n",
    "    return word_set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 将词汇表中的前N个高频词去掉\n",
    "# all_word_list: 词汇表 deleteN: N stopwords_set: 无用词汇集合\n",
    "# 该函数直接不考虑词汇表中的前0-deleteN-1个单词, 从第deleteN开始逐一判断每一个单词是否符合保留的条件\n",
    "# 符合的话就放入feature_words中, 不符合就抛弃, 生成新的词汇表\n",
    "# 返回: 新的词汇表\n",
    "def words_dict(all_word_list, deleteN, stopwords_set = set()):\n",
    "    feature_words = []\n",
    "    n = 1\n",
    "    for t in range(deleteN, len(all_word_list), 1):\n",
    "        if n > 1000:\n",
    "            break\n",
    "        if not all_word_list[t].isdigit() and all_word_list[t] not in stopwords_set and 1 < len(all_word_list[t]) < 5:\n",
    "            feature_words.append(all_word_list[t])\n",
    "        n += 1\n",
    "    return feature_words"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 文本生成特征向量\n",
    "# train_data_list, test_data_list, feature_words: 训练集, 测试集, 新词汇表\n",
    "# 遍历训练集中的每一个文档\n",
    "#    遍历词汇表中的每一个单词, 如果这个单词出现在文档中, 就把这个位置设置为1, 否则为0\n",
    "#    所以每一个单词生成的词向量长度是len(feature_words)\n",
    "# 返回: 训练集词向量, 测试集词向量\n",
    "def TextFeatures(train_data_list, test_data_list, feature_words):\n",
    "    def text_feature(text, feature_words):\n",
    "        text_words = set(text)\n",
    "        features = [1 if word in text_words else 0 for word in feature_words]\n",
    "        return features\n",
    "    train_feature_list = [text_feature(text, feature_words) for text in train_data_list]\n",
    "    test_feature_list = [text_feature(text, feature_words) for text in test_data_list]\n",
    "    return train_feature_list, test_feature_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 制作分类器\n",
    "def TextClassifier(train_feature_list, test_feature_list, train_class_list, test_class_list):\n",
    "    classifier = MultinomialNB().fit(train_feature_list, train_class_list)\n",
    "    test_accuracy = classifier.score(test_feature_list, test_class_list)\n",
    "    return test_accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEWCAYAAAB1xKBvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nc4l4wHDOOZfI/wdiSwEQRrXuJwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f1a26c645f8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if __name__ == '__main__':\n",
    "    folder_path = './Machine-Learning-master/Naive Bayes/SogouC/Sample'\n",
    "    all_word_list, train_data_list, test_data_list, train_class_list, test_class_list = TextProcessing(folder_path)\n",
    "    stopwords_file = './Machine-Learning-master/Naive Bayes/stopwords_cn.txt'\n",
    "    stopwords_set = makeWordSet(stopwords_file)\n",
    "#     feature_words = words_dict(all_word_list, 100, stopwords_set)\n",
    "    test_accuracy_list = []\n",
    "    deleteNs = range(0, 1000, 20)\n",
    "    for deleteN in deleteNs:\n",
    "        feature_words = words_dict(all_word_list, deleteN, stopwords_set)\n",
    "        train_feature_list, test_feature_list = TextFeatures(train_data_list, test_data_list, feature_words)\n",
    "        test_accuracy = TextClassifier(train_feature_list, test_feature_list,train_class_list, test_class_list)\n",
    "        test_accuracy_list.append(test_accuracy)\n",
    "    plt.figure()\n",
    "    plt.title('Relationship of deleteNs and test_accuracy')\n",
    "    plt.plot(deleteNs, test_accuracy_list)\n",
    "    plt.xlabel('deleteNs')\n",
    "    plt.ylabel('test_accuracy')\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.526315789474\n",
      "0.526315789474 [0.52631578947368418]\n"
     ]
    }
   ],
   "source": [
    "if __name__ == '__main__':\n",
    "    folder_path = './Machine-Learning-master/Naive Bayes/SogouC/Sample'\n",
    "    all_word_list, train_data_list, test_data_list, train_class_list, test_class_list = TextProcessing(folder_path)\n",
    "    stopwords_file = './Machine-Learning-master/Naive Bayes/stopwords_cn.txt'\n",
    "    stopwords_set = makeWordSet(stopwords_file)\n",
    "    \n",
    "    test_accuracy_list = []\n",
    "    feature_words = words_dict(all_word_list, 300, stopwords_set)\n",
    "    train_feature_list, test_feature_list = TextFeatures(train_data_list, test_data_list, feature_words)\n",
    "    test_accuracy = TextClassifier(train_feature_list, test_feature_list, train_class_list, test_class_list)\n",
    "    test_accuracy_list.append(test_accuracy)\n",
    "    ave = lambda c: sum(c) / len(c)\n",
    "    \n",
    "    print(ave(test_accuracy_list))\n",
    "    print(test_accuracy, test_accuracy_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}