text_classification_keras_glove_2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "数据量:\n",
      " 18846\n",
      "数据集:\n",
      " From: mblawson@midway.ecn.uoknor.edu (Matthew B Lawson)\n",
      "Subject: Which high-performance VLB video card?\n",
      "Summary: Seek recommendations for VLB video card\n",
      "Nntp-Posting-Host: midway.ecn.uoknor.edu\n",
      "Organization: Engineering Computer Network, University of Oklahoma, Norman, OK, USA\n",
      "Keywords: orchid, stealth, vlb\n",
      "Lines: 21\n",
      "\n",
      "  My brother is in the market for a high-performance video card that supports\n",
      "VESA local bus with 1-2MB RAM.  Does anyone have suggestions/ideas on:\n",
      "\n",
      "  - Diamond Stealth Pro Local Bus\n",
      "\n",
      "  - Orchid Farenheit 1280\n",
      "\n",
      "  - ATI Graphics Ultra Pro\n",
      "\n",
      "  - Any other high-performance VLB card\n",
      "\n",
      "\n",
      "Please post or email.  Thank you!\n",
      "\n",
      "  - Matt\n",
      "\n",
      "-- \n",
      "    |  Matthew B. Lawson <------------> (mblawson@essex.ecn.uoknor.edu)  |   \n",
      "  --+-- \"Now I, Nebuchadnezzar, praise and exalt and glorify the King  --+-- \n",
      "    |   of heaven, because everything he does is right and all his ways  |   \n",
      "    |   are just.\" - Nebuchadnezzar, king of Babylon, 562 B.C.           |   \n",
      "\n"
     ]
    }
   ],
   "source": [
    "# 获取所有的数据\n",
    "# 不做标题,脚注,引用处理\n",
    "from sklearn.datasets import fetch_20newsgroups\n",
    "newsgroups_all = fetch_20newsgroups(subset='all')\n",
    "\n",
    "all_data = newsgroups_all.data\n",
    "# 查看数据集和数据量\n",
    "print(\"数据量:\\n\",len(all_data))\n",
    "print(\"数据集:\\n\",all_data[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "数据量:\n",
      " 18846\n",
      "数据集:\n",
      " from mblawson midway ecn uoknor edu matthew b lawson subject which high performance vlb video card summary seek recommendations for vlb video card nntp posting host midway ecn uoknor edu organization engineering computer network university of oklahoma norman ok usa keywords orchid stealth vlb lines 21 my brother is in the market for a high performance video card that supports vesa local bus with 1 2mb ram does anyone have suggestions ideas on diamond stealth pro local bus orchid farenheit 1280 ati graphics ultra pro any other high performance vlb card please post or email thank you matt matthew b lawson mblawson essex ecn uoknor edu now i nebuchadnezzar praise and exalt and glorify the king of heaven because everything he does is right and all his ways are just nebuchadnezzar king of babylon 562 b c\n"
     ]
    }
   ],
   "source": [
    "# 去除符号，处理大小写问题\n",
    "import re\n",
    "def clear_data(data):\n",
    "    res = re.sub('[\\W]+', ' ', data.lower())\n",
    "    return res.strip()\n",
    "all_data=list(map(clear_data,all_data))\n",
    "# 查看数据集和数据量\n",
    "print(\"数据量:\\n\",len(all_data))\n",
    "print(\"数据集:\\n\",all_data[1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "标签名:\n",
      " ['alt.atheism', 'comp.graphics', 'comp.os.ms-windows.misc', 'comp.sys.ibm.pc.hardware', 'comp.sys.mac.hardware', 'comp.windows.x', 'misc.forsale', 'rec.autos', 'rec.motorcycles', 'rec.sport.baseball', 'rec.sport.hockey', 'sci.crypt', 'sci.electronics', 'sci.med', 'sci.space', 'soc.religion.christian', 'talk.politics.guns', 'talk.politics.mideast', 'talk.politics.misc', 'talk.religion.misc']\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Text(0.5,1,'All Data Distribution')"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a100c49b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from collections import Counter\n",
    "%matplotlib inline\n",
    "labels = newsgroups_all.target\n",
    "label_names = newsgroups_all.target_names\n",
    "# 查看标签名\n",
    "print(\"标签名:\\n\",label_names)\n",
    "# 查看数据的分布情况\n",
    "label_des = Counter(labels)\n",
    "plt.figure(figsize=(20,6))\n",
    "plt.bar(list(label_des.keys()), list(label_des.values()), tick_label = label_names,width = 0.6)\n",
    "plt.xlabel('Categories')\n",
    "plt.ylabel('Frequency')\n",
    "plt.title('All Data Distribution')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[('the', 240794), ('to', 124248), ('of', 114080), ('a', 105456), ('and', 95818), ('i', 87457), ('in', 81519), ('is', 70316), ('that', 65129), ('ax', 62416)]\n"
     ]
    }
   ],
   "source": [
    "# 统计词频和文章长度\n",
    "total_word = Counter()\n",
    "d_length = []\n",
    "def count_doc(data):\n",
    "    for d in data:\n",
    "        # 处理后，获取这篇文档的所有单词数\n",
    "        d_length.append(len(d.split()))\n",
    "        for word in d.split():\n",
    "            # 统计每个单词出现的次数\n",
    "            total_word[word] += 1\n",
    "    return d_length, total_word\n",
    "\n",
    "d_length, total_word = count_doc(all_data)\n",
    "print(total_word.most_common(10))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a110a9be0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 文档词数分布情况\n",
    "plt.hist(d_length, range=(0,1200),bins=20)  \n",
    "plt.xlabel('Document Words Count')\n",
    "plt.ylabel('Frequency')\n",
    "plt.title('Document Words Count')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a110903c8>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 文档长度分布\n",
    "x_val = [i[0] for i in total_word.most_common(20)]\n",
    "y_val = [i[1] for i in total_word.most_common(20)]\n",
    "plt.bar(x=range(20),height=y_val,width=0.5)\n",
    "plt.xticks(range(20),x_val)\n",
    "plt.title('Most Frequent Words')\n",
    "plt.ylabel('Frequency')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from keras.preprocessing.text import Tokenizer\n",
    "from keras.preprocessing.sequence import pad_sequences\n",
    "\n",
    "# 保留词频小于等于20000个单词\n",
    "MAX_NB_WORDS = 20000 \n",
    "# 每篇文档限定1000词以内\n",
    "MAX_SEQUENCE_LENGTH = 1000 \n",
    "\n",
    "# 构建分词器\n",
    "tokenizer = Tokenizer(num_words=MAX_NB_WORDS)\n",
    "# 将所有数据放到分词器里边\n",
    "tokenizer.fit_on_texts(all_data)\n",
    "# 文本转化为数字序列\n",
    "sequences = tokenizer.texts_to_sequences(all_data)\n",
    "# 构建词汇表\n",
    "word_index = tokenizer.word_index \n",
    "\n",
    "# 按照最大文本长度截断文本\n",
    "features = pad_sequences(sequences, maxlen=MAX_SEQUENCE_LENGTH)\n",
    "# 查看数字序列和词汇表\n",
    "# print(\"数字序列:\\n\",sequences[:2])\n",
    "# print(\"词汇表:\\n\",len(word_index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "经独热编码之后的标签: [[0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0.]\n",
      " [0. 0. 0. 1. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0. 0.]]\n"
     ]
    }
   ],
   "source": [
    "# 对标签进行独热编码\n",
    "from keras.utils import to_categorical\n",
    "labels = to_categorical(labels)\n",
    "print(\"经独热编码之后的标签:\",labels[:2])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "x的维度:\n",
      " (12060, 1000) (3016, 1000) (3770, 1000)\n",
      "y的维度:\n",
      " (12060, 20) (3016, 20) (3770, 20)\n"
     ]
    }
   ],
   "source": [
    "# 将数据拆分成训练集,验证集,测试集\n",
    "from sklearn.model_selection import train_test_split\n",
    "# 测试集占据0.2\n",
    "x_tv, x_test, y_tv, y_test = train_test_split(features, labels, test_size=0.2) \n",
    "# 验证集占据训练集中的0.2\n",
    "x_train, x_val, y_train, y_val = train_test_split(x_tv, y_tv, test_size=0.2) \n",
    "print(\"x的维度:\\n\",x_train.shape, x_val.shape, x_test.shape)\n",
    "print(\"y的维度:\\n\",y_train.shape, y_val.shape, y_test.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "embeddings_index len:\n",
      " 400000\n",
      "embeddings_index val:\n",
      " [-0.038194 -0.24487   0.72812  -0.39961   0.083172  0.043953 -0.39141\n",
      "  0.3344   -0.57545   0.087459  0.28787  -0.06731   0.30906  -0.26384\n",
      " -0.13231  -0.20757   0.33395  -0.33848  -0.31743  -0.48336   0.1464\n",
      " -0.37304   0.34577   0.052041  0.44946  -0.46971   0.02628  -0.54155\n",
      " -0.15518  -0.14107  -0.039722  0.28277   0.14393   0.23464  -0.31021\n",
      "  0.086173  0.20397   0.52624   0.17164  -0.082378 -0.71787  -0.41531\n",
      "  0.20335  -0.12763   0.41367   0.55187   0.57908  -0.33477  -0.36559\n",
      " -0.54857  -0.062892  0.26584   0.30205   0.99775  -0.80481  -3.0243\n",
      "  0.01254  -0.36942   2.2167    0.72201  -0.24978   0.92136   0.034514\n",
      "  0.46745   1.1079   -0.19358  -0.074575  0.23353  -0.052062 -0.22044\n",
      "  0.057162 -0.15806  -0.30798  -0.41625   0.37972   0.15006  -0.53212\n",
      " -0.2055   -1.2526    0.071624  0.70565   0.49744  -0.42063   0.26148\n",
      " -1.538    -0.30223  -0.073438 -0.28312   0.37104  -0.25217   0.016215\n",
      " -0.017099 -0.38984   0.87424  -0.72569  -0.51058  -0.52028  -0.1459\n",
      "  0.8278    0.27062 ]\n"
     ]
    }
   ],
   "source": [
    "# 从GloVe文件中解析出每个词和它所对应的词向量，并用字典的方式存储\n",
    "import numpy as np\n",
    "embeddings_index = {}\n",
    "with open('glove.6B.100d.txt', 'r', encoding='utf-8') as txtfile:\n",
    "    lines = txtfile.readlines()\n",
    "    for line in lines:\n",
    "        values = line.split()\n",
    "        word = values[0]\n",
    "        coefs = np.asarray(values[1:], dtype='float32')\n",
    "        embeddings_index[word] = coefs\n",
    "\n",
    "print(\"embeddings_index len:\\n\",len(embeddings_index))\n",
    "print(\"embeddings_index val:\\n\",embeddings_index['the'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "embedding_matrix val:\n",
      " [[ 0.          0.          0.         ...  0.          0.\n",
      "   0.        ]\n",
      " [-0.038194   -0.24487001  0.72812003 ... -0.1459      0.82779998\n",
      "   0.27061999]\n",
      " [-0.18970001  0.050024    0.19084001 ... -0.39804     0.47646999\n",
      "  -0.15983   ]\n",
      " ...\n",
      " [ 0.          0.          0.         ...  0.          0.\n",
      "   0.        ]\n",
      " [ 0.          0.          0.         ...  0.          0.\n",
      "   0.        ]\n",
      " [ 0.          0.          0.         ...  0.          0.\n",
      "   0.        ]]\n",
      "零值比列：\n",
      " 0.5953664240855722\n"
     ]
    }
   ],
   "source": [
    "# 构建词向量矩阵, 没有的单词的词向量均为0\n",
    "embedding_dim = 100\n",
    "embedding_matrix = np.zeros((len(word_index) + 1, embedding_dim))\n",
    "\n",
    "# 统计零值\n",
    "count_zero = 0\n",
    "for word, idx in word_index.items():\n",
    "    embed_vector = embeddings_index.get(word)\n",
    "    if embed_vector is not None:\n",
    "        count_zero += 1\n",
    "        embedding_matrix[idx] = embed_vector\n",
    "        \n",
    "print(\"embedding_matrix val:\\n\",embedding_matrix)\n",
    "prop = count_zero * 1.0 / embedding_matrix.shape[0]\n",
    "print(\"零值比列：\\n\",1 - prop)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# 词向量加载到embedding层中\n",
    "from keras.layers import Embedding\n",
    "embedding_layer = Embedding(len(word_index) + 1,\n",
    "                            embedding_dim,\n",
    "                            weights=[embedding_matrix],\n",
    "                            input_length=MAX_SEQUENCE_LENGTH,\n",
    "                            trainable=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# 构建CNN模型"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import keras\n",
    "from keras.layers import Dense,Input\n",
    "from keras.layers import Conv1D, MaxPooling1D, Concatenate, Flatten, Dropout\n",
    "from keras.models import Model\n",
    "tb = [keras.callbacks.TensorBoard(log_dir='./logs')]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_2 (InputLayer)         (None, 1000)              0         \n",
      "_________________________________________________________________\n",
      "embedding_4 (Embedding)      (None, 1000, 100)         16846600  \n",
      "_________________________________________________________________\n",
      "conv1d_4 (Conv1D)            (None, 996, 128)          64128     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_4 (MaxPooling1 (None, 199, 128)          0         \n",
      "_________________________________________________________________\n",
      "conv1d_5 (Conv1D)            (None, 195, 128)          82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_5 (MaxPooling1 (None, 39, 128)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_6 (Conv1D)            (None, 35, 128)           82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_6 (MaxPooling1 (None, 1, 128)            0         \n",
      "_________________________________________________________________\n",
      "flatten_2 (Flatten)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_2 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_3 (Dense)              (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dense_4 (Dense)              (None, 20)                2580      \n",
      "=================================================================\n",
      "Total params: 17,093,916\n",
      "Trainable params: 17,093,916\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 12060 samples, validate on 3016 samples\n",
      "Epoch 1/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 2.6701 - acc: 0.1319 - val_loss: 2.0623 - val_acc: 0.2775\n",
      "Epoch 2/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 1.6805 - acc: 0.3890 - val_loss: 1.2755 - val_acc: 0.5458\n",
      "Epoch 3/25\n",
      "12060/12060 [==============================] - 154s 13ms/step - loss: 1.0466 - acc: 0.6135 - val_loss: 0.9793 - val_acc: 0.6565\n",
      "Epoch 4/25\n",
      "12060/12060 [==============================] - 142s 12ms/step - loss: 0.6663 - acc: 0.7637 - val_loss: 0.7838 - val_acc: 0.7420\n",
      "Epoch 5/25\n",
      "12060/12060 [==============================] - 145s 12ms/step - loss: 0.4284 - acc: 0.8556 - val_loss: 0.6524 - val_acc: 0.7991\n",
      "Epoch 6/25\n",
      "12060/12060 [==============================] - 149s 12ms/step - loss: 0.2674 - acc: 0.9153 - val_loss: 0.6344 - val_acc: 0.8193\n",
      "Epoch 7/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.1684 - acc: 0.9480 - val_loss: 0.6070 - val_acc: 0.8345\n",
      "Epoch 8/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.1031 - acc: 0.9702 - val_loss: 0.6637 - val_acc: 0.8299\n",
      "Epoch 9/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.0707 - acc: 0.9782 - val_loss: 0.7126 - val_acc: 0.8349\n",
      "Epoch 10/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.0447 - acc: 0.9871 - val_loss: 0.7602 - val_acc: 0.8369\n",
      "Epoch 11/25\n",
      "12060/12060 [==============================] - 155s 13ms/step - loss: 0.0358 - acc: 0.9894 - val_loss: 0.7537 - val_acc: 0.8455\n",
      "Epoch 12/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.0280 - acc: 0.9926 - val_loss: 0.8043 - val_acc: 0.8465\n",
      "Epoch 13/25\n",
      "12060/12060 [==============================] - 143s 12ms/step - loss: 0.0273 - acc: 0.9922 - val_loss: 0.7525 - val_acc: 0.8531\n",
      "Epoch 14/25\n",
      "12060/12060 [==============================] - 141s 12ms/step - loss: 0.0230 - acc: 0.9936 - val_loss: 0.9128 - val_acc: 0.8339\n",
      "Epoch 15/25\n",
      "12060/12060 [==============================] - 151s 12ms/step - loss: 0.0251 - acc: 0.9924 - val_loss: 0.8462 - val_acc: 0.8481\n",
      "Epoch 16/25\n",
      "12060/12060 [==============================] - 155s 13ms/step - loss: 0.0245 - acc: 0.9933 - val_loss: 0.8964 - val_acc: 0.8418\n",
      "Epoch 17/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.0179 - acc: 0.9948 - val_loss: 0.8303 - val_acc: 0.8534\n",
      "Epoch 18/25\n",
      "12060/12060 [==============================] - 151s 12ms/step - loss: 0.0192 - acc: 0.9951 - val_loss: 0.8189 - val_acc: 0.8591\n",
      "Epoch 19/25\n",
      "12060/12060 [==============================] - 152s 13ms/step - loss: 0.0207 - acc: 0.9936 - val_loss: 0.7616 - val_acc: 0.8654\n",
      "Epoch 20/25\n",
      "12060/12060 [==============================] - 150s 12ms/step - loss: 0.0158 - acc: 0.9949 - val_loss: 0.7761 - val_acc: 0.8641\n",
      "Epoch 21/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0143 - acc: 0.9963 - val_loss: 0.8340 - val_acc: 0.8684\n",
      "Epoch 22/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0153 - acc: 0.9954 - val_loss: 0.8374 - val_acc: 0.8667\n",
      "Epoch 23/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0098 - acc: 0.9973 - val_loss: 0.8180 - val_acc: 0.8717\n",
      "Epoch 24/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.0124 - acc: 0.9963 - val_loss: 0.8331 - val_acc: 0.8667\n",
      "Epoch 25/25\n",
      "12060/12060 [==============================] - 155s 13ms/step - loss: 0.0073 - acc: 0.9983 - val_loss: 0.8801 - val_acc: 0.8614\n"
     ]
    }
   ],
   "source": [
    "# （Conv1D-MaxPooling1D）* 3 - Flatten - Dropout - Dense * 2\n",
    "main_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='float64')\n",
    "# 使用已训练完成的词向量\n",
    "embedding_layer = Embedding(len(word_index) + 1, embedding_dim,\n",
    "                            weights=[embedding_matrix],\n",
    "                            input_length=MAX_SEQUENCE_LENGTH,\n",
    "                            trainable=True)\n",
    "embed = embedding_layer(main_input)\n",
    "# 使用3层卷积池化\n",
    "x = Conv1D(128, 5, activation='relu')(embed)\n",
    "x = MaxPooling1D(5)(x)\n",
    "x = Conv1D(128, 5, activation='relu')(x)\n",
    "x = MaxPooling1D(5)(x)\n",
    "x = Conv1D(128, 5, activation='relu')(x)\n",
    "x = MaxPooling1D(35)(x)\n",
    "x = Flatten()(x)\n",
    "x = Dropout(0.2)(x) # 丢弃20%\n",
    "x = Dense(128, activation='relu')(x)\n",
    "preds = Dense(20, activation='softmax')(x)\n",
    "model = Model(inputs=main_input, outputs=preds)\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "print(model.summary()) # 查看模型概况\n",
    "\n",
    "# history_cnn记录训练结果\n",
    "history_cnn = model.fit(x_train, y_train, batch_size=128, epochs=25, callbacks=tb, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a27839828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 绘制模型的训练和验证曲线\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "plt.plot(history_cnn.history['acc'])\n",
    "plt.plot(history_cnn.history['val_acc']) \n",
    "\n",
    "plt.legend(['Train line', 'Val line'], loc='lower right', fontsize=16)\n",
    "\n",
    "plt.title('CNN[(Conv1D-MaxPooling1D)* 3 - Flatten - Dropout - Dense * 2] Correct Rate', fontsize=20)\n",
    "plt.xlabel('Epoch Count', fontsize=18)\n",
    "plt.ylabel('Correct Rate', fontsize=18)\n",
    "\n",
    "plt.xticks(range(0, 26, 5), fontsize=16)\n",
    "plt.yticks(fontsize=16)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3770/3770 [==============================] - 16s 4ms/step\n",
      "Test score: 0.7753802068195546\n",
      "Test accuracy: 0.8769230766701129\n"
     ]
    }
   ],
   "source": [
    "# 模型测试 评估\n",
    "score, acc = model.evaluate(x_test, y_test, batch_size=128)\n",
    "print('Test score:', score)\n",
    "print('Test accuracy:', acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "input_3 (InputLayer)         (None, 1000)              0         \n",
      "_________________________________________________________________\n",
      "embedding_5 (Embedding)      (None, 1000, 100)         16846600  \n",
      "_________________________________________________________________\n",
      "conv1d_7 (Conv1D)            (None, 996, 128)          64128     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_7 (MaxPooling1 (None, 199, 128)          0         \n",
      "_________________________________________________________________\n",
      "dropout_3 (Dropout)          (None, 199, 128)          0         \n",
      "_________________________________________________________________\n",
      "conv1d_8 (Conv1D)            (None, 195, 128)          82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_8 (MaxPooling1 (None, 39, 128)           0         \n",
      "_________________________________________________________________\n",
      "dropout_4 (Dropout)          (None, 39, 128)           0         \n",
      "_________________________________________________________________\n",
      "conv1d_9 (Conv1D)            (None, 35, 128)           82048     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_9 (MaxPooling1 (None, 1, 128)            0         \n",
      "_________________________________________________________________\n",
      "flatten_3 (Flatten)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dropout_5 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_5 (Dense)              (None, 128)               16512     \n",
      "_________________________________________________________________\n",
      "dropout_6 (Dropout)          (None, 128)               0         \n",
      "_________________________________________________________________\n",
      "dense_6 (Dense)              (None, 20)                2580      \n",
      "=================================================================\n",
      "Total params: 17,093,916\n",
      "Trainable params: 17,093,916\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 12060 samples, validate on 3016 samples\n",
      "Epoch 1/25\n",
      "12060/12060 [==============================] - 153s 13ms/step - loss: 2.7624 - acc: 0.1146 - val_loss: 2.2269 - val_acc: 0.2424\n",
      "Epoch 2/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 2.0609 - acc: 0.2635 - val_loss: 1.7374 - val_acc: 0.3810\n",
      "Epoch 3/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 1.4750 - acc: 0.4546 - val_loss: 1.1495 - val_acc: 0.5938\n",
      "Epoch 4/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.9780 - acc: 0.6461 - val_loss: 0.7946 - val_acc: 0.7470\n",
      "Epoch 5/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.6685 - acc: 0.7701 - val_loss: 0.6586 - val_acc: 0.7934\n",
      "Epoch 6/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.4700 - acc: 0.8461 - val_loss: 0.5715 - val_acc: 0.8236\n",
      "Epoch 7/25\n",
      "12060/12060 [==============================] - 148s 12ms/step - loss: 0.3350 - acc: 0.8910 - val_loss: 0.5421 - val_acc: 0.8408\n",
      "Epoch 8/25\n",
      "12060/12060 [==============================] - 156s 13ms/step - loss: 0.2534 - acc: 0.9157 - val_loss: 0.5338 - val_acc: 0.8548\n",
      "Epoch 9/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.1787 - acc: 0.9401 - val_loss: 0.5588 - val_acc: 0.8578\n",
      "Epoch 10/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.1431 - acc: 0.9540 - val_loss: 0.5956 - val_acc: 0.8534\n",
      "Epoch 11/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.1068 - acc: 0.9657 - val_loss: 0.5978 - val_acc: 0.8594\n",
      "Epoch 12/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0914 - acc: 0.9700 - val_loss: 0.6295 - val_acc: 0.8604\n",
      "Epoch 13/25\n",
      "12060/12060 [==============================] - 158s 13ms/step - loss: 0.0823 - acc: 0.9748 - val_loss: 0.6043 - val_acc: 0.8730\n",
      "Epoch 14/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0623 - acc: 0.9805 - val_loss: 0.6366 - val_acc: 0.8740\n",
      "Epoch 15/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0570 - acc: 0.9821 - val_loss: 0.6543 - val_acc: 0.8747\n",
      "Epoch 16/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0526 - acc: 0.9824 - val_loss: 0.7707 - val_acc: 0.8631\n",
      "Epoch 17/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0505 - acc: 0.9832 - val_loss: 0.6262 - val_acc: 0.8757\n",
      "Epoch 18/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0337 - acc: 0.9883 - val_loss: 0.6837 - val_acc: 0.8697\n",
      "Epoch 19/25\n",
      "12060/12060 [==============================] - 146s 12ms/step - loss: 0.0326 - acc: 0.9900 - val_loss: 0.7313 - val_acc: 0.8740\n",
      "Epoch 20/25\n",
      "12060/12060 [==============================] - 158s 13ms/step - loss: 0.0347 - acc: 0.9896 - val_loss: 0.7064 - val_acc: 0.8714\n",
      "Epoch 21/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0358 - acc: 0.9896 - val_loss: 0.6942 - val_acc: 0.8820\n",
      "Epoch 22/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0230 - acc: 0.9940 - val_loss: 0.7032 - val_acc: 0.8826\n",
      "Epoch 23/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0302 - acc: 0.9913 - val_loss: 0.6929 - val_acc: 0.8800\n",
      "Epoch 24/25\n",
      "12060/12060 [==============================] - 147s 12ms/step - loss: 0.0284 - acc: 0.9922 - val_loss: 0.7116 - val_acc: 0.8813\n",
      "Epoch 25/25\n",
      "12060/12060 [==============================] - 151s 12ms/step - loss: 0.0166 - acc: 0.9952 - val_loss: 0.7553 - val_acc: 0.8793\n"
     ]
    }
   ],
   "source": [
    "# (Conv1D-MaxPooling1D-Dropout)* 2-Conv1D-MaxPooling1D-Dropout-Dense-Dropout-Dense\n",
    "main_input = Input(shape=(MAX_SEQUENCE_LENGTH,), dtype='float64')\n",
    "# 使用已训练完成的词向量\n",
    "embedding_layer = Embedding(len(word_index) + 1, embedding_dim,\n",
    "                            weights=[embedding_matrix],\n",
    "                            input_length=MAX_SEQUENCE_LENGTH,\n",
    "                            trainable=True)\n",
    "embed = embedding_layer(main_input)\n",
    "# 使用3层卷积池化,每层之间使用Dropout来防止过拟合\n",
    "x = Conv1D(128, 5, activation='relu')(embed)\n",
    "x = MaxPooling1D(5)(x)\n",
    "x = Dropout(0.2)(x)\n",
    "x = Conv1D(128, 5, activation='relu')(x)\n",
    "x = MaxPooling1D(5)(x)\n",
    "x = Dropout(0.2)(x)\n",
    "x = Conv1D(128, 5, activation='relu')(x)\n",
    "x = MaxPooling1D(35)(x)\n",
    "x = Flatten()(x)\n",
    "x = Dropout(0.2)(x)\n",
    "x = Dense(128, activation='relu')(x)\n",
    "x = Dropout(0.2)(x)\n",
    "preds = Dense(20, activation='softmax')(x)\n",
    "model = Model(inputs=main_input, outputs=preds)\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "print(model.summary()) # 查看模型概况\n",
    "\n",
    "# history_cnn记录训练结果\n",
    "history_cnn = model.fit(x_train, y_train, batch_size=128, epochs=25, callbacks=tb, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a3a05cb38>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 绘制模型的训练和验证曲线\n",
    "plt.style.use('ggplot')\n",
    "\n",
    "plt.plot(history_cnn.history['acc'])\n",
    "plt.plot(history_cnn.history['val_acc']) \n",
    "\n",
    "plt.legend(['Train line', 'Val line'], loc='lower right', fontsize=16)\n",
    "\n",
    "plt.title('CNN[(Conv1D-MaxPooling1D-Dropout)* 2-Conv1D-MaxPooling1D-Dropout-Dense-Dropout-Dense] Correct Rate', fontsize=20)\n",
    "plt.xlabel('Epoch Count', fontsize=18)\n",
    "plt.ylabel('Correct Rate', fontsize=18)\n",
    "\n",
    "plt.xticks(range(0, 26, 5), fontsize=16)\n",
    "plt.yticks(fontsize=16)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3770/3770 [==============================] - 14s 4ms/step\n",
      "Test score: 0.7242819279511348\n",
      "Test accuracy: 0.8843501329738203\n"
     ]
    }
   ],
   "source": [
    "# 模型测试 评估\n",
    "score, acc = model.evaluate(x_test, y_test, batch_size=128)\n",
    "print('Test score:', score)\n",
    "print('Test accuracy:', acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_6 (Embedding)      (None, 1000, 100)         16846600  \n",
      "_________________________________________________________________\n",
      "lstm_1 (LSTM)                (None, 100)               80400     \n",
      "_________________________________________________________________\n",
      "dense_7 (Dense)              (None, 20)                2020      \n",
      "=================================================================\n",
      "Total params: 16,929,020\n",
      "Trainable params: 16,929,020\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 12060 samples, validate on 3016 samples\n",
      "Epoch 1/60\n",
      "12060/12060 [==============================] - 265s 22ms/step - loss: 3.1980 - acc: 0.1274 - val_loss: 2.8544 - val_acc: 0.1940\n",
      "Epoch 2/60\n",
      "12060/12060 [==============================] - 260s 22ms/step - loss: 2.7549 - acc: 0.2108 - val_loss: 2.5377 - val_acc: 0.2304\n",
      "Epoch 3/60\n",
      "12060/12060 [==============================] - 257s 21ms/step - loss: 2.5181 - acc: 0.2580 - val_loss: 2.3312 - val_acc: 0.3054\n",
      "Epoch 4/60\n",
      "12060/12060 [==============================] - 256s 21ms/step - loss: 2.3201 - acc: 0.3219 - val_loss: 2.1575 - val_acc: 0.3710\n",
      "Epoch 5/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 2.0445 - acc: 0.3990 - val_loss: 1.8658 - val_acc: 0.4532\n",
      "Epoch 6/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 1.7939 - acc: 0.4892 - val_loss: 1.6342 - val_acc: 0.5716\n",
      "Epoch 7/60\n",
      "12060/12060 [==============================] - 252s 21ms/step - loss: 1.5831 - acc: 0.5738 - val_loss: 1.4558 - val_acc: 0.6267\n",
      "Epoch 8/60\n",
      "12060/12060 [==============================] - 257s 21ms/step - loss: 1.4068 - acc: 0.6364 - val_loss: 1.3235 - val_acc: 0.6757\n",
      "Epoch 9/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 1.2603 - acc: 0.6857 - val_loss: 1.2411 - val_acc: 0.7029\n",
      "Epoch 10/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 1.1679 - acc: 0.7228 - val_loss: 1.1634 - val_acc: 0.7175\n",
      "Epoch 11/60\n",
      "12060/12060 [==============================] - 253s 21ms/step - loss: 1.0780 - acc: 0.7510 - val_loss: 1.1046 - val_acc: 0.7490\n",
      "Epoch 12/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 1.0036 - acc: 0.7820 - val_loss: 1.0451 - val_acc: 0.7716\n",
      "Epoch 13/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.9316 - acc: 0.8100 - val_loss: 0.9992 - val_acc: 0.7934\n",
      "Epoch 14/60\n",
      "12060/12060 [==============================] - 256s 21ms/step - loss: 0.8830 - acc: 0.8297 - val_loss: 0.9816 - val_acc: 0.7944\n",
      "Epoch 15/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.8260 - acc: 0.8507 - val_loss: 0.9300 - val_acc: 0.8113\n",
      "Epoch 16/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.7845 - acc: 0.8666 - val_loss: 0.9094 - val_acc: 0.8226\n",
      "Epoch 17/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.7387 - acc: 0.8839 - val_loss: 0.8840 - val_acc: 0.8389\n",
      "Epoch 18/60\n",
      "12060/12060 [==============================] - 253s 21ms/step - loss: 0.7104 - acc: 0.8920 - val_loss: 0.9644 - val_acc: 0.8077\n",
      "Epoch 19/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.6999 - acc: 0.9002 - val_loss: 0.8540 - val_acc: 0.8515\n",
      "Epoch 20/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.6480 - acc: 0.9144 - val_loss: 0.8532 - val_acc: 0.8462\n",
      "Epoch 21/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.6247 - acc: 0.9195 - val_loss: 0.8157 - val_acc: 0.8564\n",
      "Epoch 22/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5929 - acc: 0.9332 - val_loss: 0.8155 - val_acc: 0.8568\n",
      "Epoch 23/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.5720 - acc: 0.9381 - val_loss: 0.8134 - val_acc: 0.8637\n",
      "Epoch 24/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5630 - acc: 0.9393 - val_loss: 0.8048 - val_acc: 0.8651\n",
      "Epoch 25/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5734 - acc: 0.9408 - val_loss: 0.8019 - val_acc: 0.8674\n",
      "Epoch 26/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5505 - acc: 0.9478 - val_loss: 0.7883 - val_acc: 0.8700\n",
      "Epoch 27/60\n",
      "12060/12060 [==============================] - 253s 21ms/step - loss: 0.5320 - acc: 0.9531 - val_loss: 0.8010 - val_acc: 0.8680\n",
      "Epoch 28/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5088 - acc: 0.9619 - val_loss: 0.7758 - val_acc: 0.8767\n",
      "Epoch 29/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.5066 - acc: 0.9603 - val_loss: 0.7817 - val_acc: 0.8723\n",
      "Epoch 30/60\n",
      "12060/12060 [==============================] - 255s 21ms/step - loss: 0.4876 - acc: 0.9638 - val_loss: 0.7664 - val_acc: 0.8750\n",
      "Epoch 31/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4759 - acc: 0.9676 - val_loss: 0.7569 - val_acc: 0.8810\n",
      "Epoch 32/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4623 - acc: 0.9716 - val_loss: 0.7780 - val_acc: 0.8710\n",
      "Epoch 33/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4547 - acc: 0.9731 - val_loss: 0.7554 - val_acc: 0.8826\n",
      "Epoch 34/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4478 - acc: 0.9731 - val_loss: 0.7550 - val_acc: 0.8826\n",
      "Epoch 35/60\n",
      "12060/12060 [==============================] - 253s 21ms/step - loss: 0.4389 - acc: 0.9760 - val_loss: 0.7498 - val_acc: 0.8826\n",
      "Epoch 36/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4297 - acc: 0.9789 - val_loss: 0.7605 - val_acc: 0.8846\n",
      "Epoch 37/60\n",
      "12060/12060 [==============================] - 253s 21ms/step - loss: 0.4251 - acc: 0.9776 - val_loss: 0.7509 - val_acc: 0.8813\n",
      "Epoch 38/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4271 - acc: 0.9788 - val_loss: 0.7589 - val_acc: 0.8843\n",
      "Epoch 39/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4183 - acc: 0.9803 - val_loss: 0.7434 - val_acc: 0.8853\n",
      "Epoch 40/60\n",
      "12060/12060 [==============================] - 254s 21ms/step - loss: 0.4107 - acc: 0.9824 - val_loss: 0.7528 - val_acc: 0.8863\n",
      "Epoch 41/60\n",
      "12060/12060 [==============================] - 278s 23ms/step - loss: 0.4012 - acc: 0.9842 - val_loss: 0.7545 - val_acc: 0.8833\n",
      "Epoch 42/60\n",
      "12060/12060 [==============================] - 288s 24ms/step - loss: 0.4001 - acc: 0.9835 - val_loss: 0.7428 - val_acc: 0.8876\n",
      "Epoch 43/60\n",
      "12060/12060 [==============================] - 260s 22ms/step - loss: 0.3944 - acc: 0.9847 - val_loss: 0.7456 - val_acc: 0.8863\n",
      "Epoch 44/60\n",
      "12060/12060 [==============================] - 260s 22ms/step - loss: 0.3876 - acc: 0.9862 - val_loss: 0.7441 - val_acc: 0.8853\n",
      "Epoch 45/60\n",
      "12060/12060 [==============================] - 258s 21ms/step - loss: 0.3874 - acc: 0.9851 - val_loss: 0.7297 - val_acc: 0.8879\n",
      "Epoch 46/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3827 - acc: 0.9881 - val_loss: 0.7343 - val_acc: 0.8916\n",
      "Epoch 47/60\n",
      "12060/12060 [==============================] - 259s 21ms/step - loss: 0.3779 - acc: 0.9884 - val_loss: 0.7389 - val_acc: 0.8912\n",
      "Epoch 48/60\n",
      "12060/12060 [==============================] - 258s 21ms/step - loss: 0.3731 - acc: 0.9894 - val_loss: 0.7342 - val_acc: 0.8889\n",
      "Epoch 49/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3714 - acc: 0.9892 - val_loss: 0.7300 - val_acc: 0.8899\n",
      "Epoch 50/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3673 - acc: 0.9903 - val_loss: 0.7374 - val_acc: 0.8903\n",
      "Epoch 51/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3778 - acc: 0.9878 - val_loss: 0.7256 - val_acc: 0.8869\n",
      "Epoch 52/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3742 - acc: 0.9886 - val_loss: 0.7306 - val_acc: 0.8896\n",
      "Epoch 53/60\n",
      "12060/12060 [==============================] - 262s 22ms/step - loss: 0.3701 - acc: 0.9884 - val_loss: 0.7311 - val_acc: 0.8906\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 54/60\n",
      "12060/12060 [==============================] - 261s 22ms/step - loss: 0.3622 - acc: 0.9910 - val_loss: 0.7324 - val_acc: 0.8916\n",
      "Epoch 55/60\n",
      "12060/12060 [==============================] - 260s 22ms/step - loss: 0.3646 - acc: 0.9893 - val_loss: 0.7276 - val_acc: 0.8916\n",
      "Epoch 56/60\n",
      "12060/12060 [==============================] - 263s 22ms/step - loss: 0.3606 - acc: 0.9906 - val_loss: 0.7315 - val_acc: 0.8903\n",
      "Epoch 57/60\n",
      "12060/12060 [==============================] - 259s 21ms/step - loss: 0.3515 - acc: 0.9925 - val_loss: 0.7266 - val_acc: 0.8922\n",
      "Epoch 58/60\n",
      "12060/12060 [==============================] - 262s 22ms/step - loss: 0.3498 - acc: 0.9930 - val_loss: 0.7379 - val_acc: 0.8879\n",
      "Epoch 59/60\n",
      "12060/12060 [==============================] - 260s 22ms/step - loss: 0.3475 - acc: 0.9933 - val_loss: 0.7551 - val_acc: 0.8826\n",
      "Epoch 60/60\n",
      "12060/12060 [==============================] - 262s 22ms/step - loss: 0.3550 - acc: 0.9903 - val_loss: 0.7394 - val_acc: 0.8893\n"
     ]
    }
   ],
   "source": [
    "# 模型结构：LSTM-Dense\n",
    "from keras.models import Sequential\n",
    "from keras.layers import LSTM,Activation\n",
    "from keras import regularizers\n",
    "\n",
    "model = Sequential()\n",
    "model.add(Embedding(len(word_index) + 1,embedding_dim, weights=[embedding_matrix],input_length=MAX_SEQUENCE_LENGTH, trainable=True))\n",
    "model.add(LSTM(100, dropout=0.2, recurrent_dropout=0.2))\n",
    "model.add(Dense(20, activation='softmax', kernel_regularizer=regularizers.l2(0.01), \n",
    "                activity_regularizer=regularizers.l1(0.001)))\n",
    "\n",
    "model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "print(model.summary())\n",
    "history_rnn = model.fit(x_train, y_train, batch_size=128, epochs=60, callbacks=tb, validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1a27838828>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 绘制训练和验证曲线\n",
    "plt.plot(history_rnn.history['acc'])\n",
    "plt.plot(history_rnn.history['val_acc'])\n",
    "\n",
    "plt.legend(['Train line', 'Val line'], loc='lower right', fontsize=16)\n",
    "\n",
    "plt.title('LSTM-Dense Correct Rate', fontsize=20)\n",
    "plt.xlabel('Epoch Count', fontsize=18)\n",
    "plt.ylabel('Correct Rate', fontsize=18)\n",
    "\n",
    "plt.xticks(range(0, 61, 5), fontsize=16)\n",
    "plt.yticks(fontsize=16)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3770/3770 [==============================] - 27s 7ms/step\n",
      "Test score: 0.699656054575184\n",
      "Test accuracy: 0.8949602119486275\n"
     ]
    }
   ],
   "source": [
    "# 模型测试 评估\n",
    "score, acc = model.evaluate(x_test, y_test, batch_size=128)\n",
    "print('Test score:', score)\n",
    "print('Test accuracy:', acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "_________________________________________________________________\n",
      "Layer (type)                 Output Shape              Param #   \n",
      "=================================================================\n",
      "embedding_7 (Embedding)      (None, 1000, 100)         16846600  \n",
      "_________________________________________________________________\n",
      "spatial_dropout1d_1 (Spatial (None, 1000, 100)         0         \n",
      "_________________________________________________________________\n",
      "conv1d_10 (Conv1D)           (None, 1000, 128)         64128     \n",
      "_________________________________________________________________\n",
      "max_pooling1d_10 (MaxPooling (None, 250, 128)          0         \n",
      "_________________________________________________________________\n",
      "lstm_2 (LSTM)                (None, 100)               91600     \n",
      "_________________________________________________________________\n",
      "dense_8 (Dense)              (None, 20)                2020      \n",
      "=================================================================\n",
      "Total params: 17,004,348\n",
      "Trainable params: 17,004,348\n",
      "Non-trainable params: 0\n",
      "_________________________________________________________________\n",
      "None\n",
      "Train on 12060 samples, validate on 3016 samples\n",
      "Epoch 1/60\n",
      "12060/12060 [==============================] - 280s 23ms/step - loss: 3.2244 - acc: 0.1092 - val_loss: 2.6840 - val_acc: 0.2225\n",
      "Epoch 2/60\n",
      "12060/12060 [==============================] - 221s 18ms/step - loss: 2.5900 - acc: 0.2327 - val_loss: 2.2518 - val_acc: 0.3352\n",
      "Epoch 3/60\n",
      "12060/12060 [==============================] - 201s 17ms/step - loss: 2.2327 - acc: 0.3213 - val_loss: 1.9322 - val_acc: 0.4377\n",
      "Epoch 4/60\n",
      "12060/12060 [==============================] - 194s 16ms/step - loss: 1.9626 - acc: 0.4042 - val_loss: 1.8648 - val_acc: 0.4380\n",
      "Epoch 5/60\n",
      "12060/12060 [==============================] - 195s 16ms/step - loss: 1.7615 - acc: 0.4673 - val_loss: 1.5044 - val_acc: 0.5693\n",
      "Epoch 6/60\n",
      "12060/12060 [==============================] - 196s 16ms/step - loss: 1.5676 - acc: 0.5354 - val_loss: 1.3822 - val_acc: 0.6310\n",
      "Epoch 7/60\n",
      "12060/12060 [==============================] - 195s 16ms/step - loss: 1.4332 - acc: 0.5860 - val_loss: 1.2573 - val_acc: 0.6605\n",
      "Epoch 8/60\n",
      "12060/12060 [==============================] - 195s 16ms/step - loss: 1.2977 - acc: 0.6362 - val_loss: 1.2262 - val_acc: 0.6873\n",
      "Epoch 9/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 1.2137 - acc: 0.6739 - val_loss: 1.1292 - val_acc: 0.7082\n",
      "Epoch 10/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 1.1020 - acc: 0.7182 - val_loss: 1.0520 - val_acc: 0.7384\n",
      "Epoch 11/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 1.0269 - acc: 0.7445 - val_loss: 1.0083 - val_acc: 0.7527\n",
      "Epoch 12/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.9751 - acc: 0.7607 - val_loss: 0.9848 - val_acc: 0.7676\n",
      "Epoch 13/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.9309 - acc: 0.7808 - val_loss: 0.9379 - val_acc: 0.7815\n",
      "Epoch 14/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.8854 - acc: 0.7973 - val_loss: 0.9190 - val_acc: 0.7908\n",
      "Epoch 15/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.8388 - acc: 0.8178 - val_loss: 0.8644 - val_acc: 0.8100\n",
      "Epoch 16/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.8176 - acc: 0.8298 - val_loss: 0.8430 - val_acc: 0.8206\n",
      "Epoch 17/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.7690 - acc: 0.8492 - val_loss: 0.8078 - val_acc: 0.8349\n",
      "Epoch 18/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.7354 - acc: 0.8641 - val_loss: 0.7942 - val_acc: 0.8445\n",
      "Epoch 19/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.7390 - acc: 0.8620 - val_loss: 0.8009 - val_acc: 0.8402\n",
      "Epoch 20/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.6763 - acc: 0.8857 - val_loss: 0.7661 - val_acc: 0.8521\n",
      "Epoch 21/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.6829 - acc: 0.8823 - val_loss: 0.7484 - val_acc: 0.8614\n",
      "Epoch 22/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.6304 - acc: 0.9020 - val_loss: 0.7595 - val_acc: 0.8531\n",
      "Epoch 23/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.6091 - acc: 0.9080 - val_loss: 0.7303 - val_acc: 0.8697\n",
      "Epoch 24/60\n",
      "12060/12060 [==============================] - 196s 16ms/step - loss: 0.5930 - acc: 0.9133 - val_loss: 0.7067 - val_acc: 0.8740\n",
      "Epoch 25/60\n",
      "12060/12060 [==============================] - 195s 16ms/step - loss: 0.5903 - acc: 0.9143 - val_loss: 0.7241 - val_acc: 0.8740\n",
      "Epoch 26/60\n",
      "12060/12060 [==============================] - 197s 16ms/step - loss: 0.5718 - acc: 0.9212 - val_loss: 0.6969 - val_acc: 0.8780\n",
      "Epoch 27/60\n",
      "12060/12060 [==============================] - 196s 16ms/step - loss: 0.5571 - acc: 0.9272 - val_loss: 0.6870 - val_acc: 0.8849\n",
      "Epoch 28/60\n",
      "12060/12060 [==============================] - 196s 16ms/step - loss: 0.5322 - acc: 0.9349 - val_loss: 0.6795 - val_acc: 0.8859\n",
      "Epoch 29/60\n",
      "12060/12060 [==============================] - 194s 16ms/step - loss: 0.5245 - acc: 0.9376 - val_loss: 0.6789 - val_acc: 0.8853\n",
      "Epoch 30/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.5184 - acc: 0.9368 - val_loss: 0.6781 - val_acc: 0.8846\n",
      "Epoch 31/60\n",
      "12060/12060 [==============================] - 192s 16ms/step - loss: 0.5098 - acc: 0.9441 - val_loss: 0.6719 - val_acc: 0.8859\n",
      "Epoch 32/60\n",
      "12060/12060 [==============================] - 192s 16ms/step - loss: 0.4960 - acc: 0.9457 - val_loss: 0.6637 - val_acc: 0.8922\n",
      "Epoch 33/60\n",
      "12060/12060 [==============================] - 192s 16ms/step - loss: 0.4830 - acc: 0.9505 - val_loss: 0.6550 - val_acc: 0.8939\n",
      "Epoch 34/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4771 - acc: 0.9507 - val_loss: 0.6631 - val_acc: 0.8896\n",
      "Epoch 35/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4717 - acc: 0.9561 - val_loss: 0.6755 - val_acc: 0.8856\n",
      "Epoch 36/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4583 - acc: 0.9590 - val_loss: 0.6558 - val_acc: 0.8942\n",
      "Epoch 37/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4516 - acc: 0.9585 - val_loss: 0.6558 - val_acc: 0.8942\n",
      "Epoch 38/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4442 - acc: 0.9615 - val_loss: 0.6551 - val_acc: 0.8966\n",
      "Epoch 39/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4485 - acc: 0.9600 - val_loss: 0.6467 - val_acc: 0.8939\n",
      "Epoch 40/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4274 - acc: 0.9669 - val_loss: 0.6436 - val_acc: 0.8995\n",
      "Epoch 41/60\n",
      "12060/12060 [==============================] - 192s 16ms/step - loss: 0.4222 - acc: 0.9672 - val_loss: 0.6545 - val_acc: 0.8962\n",
      "Epoch 42/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4229 - acc: 0.9689 - val_loss: 0.6510 - val_acc: 0.8932\n",
      "Epoch 43/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4125 - acc: 0.9706 - val_loss: 0.6476 - val_acc: 0.8985\n",
      "Epoch 44/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4154 - acc: 0.9696 - val_loss: 0.6422 - val_acc: 0.8982\n",
      "Epoch 45/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4128 - acc: 0.9708 - val_loss: 0.6468 - val_acc: 0.9002\n",
      "Epoch 46/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4070 - acc: 0.9730 - val_loss: 0.6467 - val_acc: 0.9022\n",
      "Epoch 47/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3962 - acc: 0.9755 - val_loss: 0.6393 - val_acc: 0.9019\n",
      "Epoch 48/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3946 - acc: 0.9761 - val_loss: 0.6403 - val_acc: 0.9029\n",
      "Epoch 49/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3960 - acc: 0.9762 - val_loss: 0.6361 - val_acc: 0.9025\n",
      "Epoch 50/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3921 - acc: 0.9739 - val_loss: 0.6385 - val_acc: 0.8982\n",
      "Epoch 51/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3829 - acc: 0.9762 - val_loss: 0.6431 - val_acc: 0.8979\n",
      "Epoch 52/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.4021 - acc: 0.9727 - val_loss: 0.6478 - val_acc: 0.8962\n",
      "Epoch 53/60\n",
      "12060/12060 [==============================] - 192s 16ms/step - loss: 0.3902 - acc: 0.9782 - val_loss: 0.6578 - val_acc: 0.8995\n",
      "Epoch 54/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3870 - acc: 0.9772 - val_loss: 0.6377 - val_acc: 0.9022\n",
      "Epoch 55/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3769 - acc: 0.9801 - val_loss: 0.6345 - val_acc: 0.9035\n",
      "Epoch 56/60\n",
      "12060/12060 [==============================] - 193s 16ms/step - loss: 0.3680 - acc: 0.9812 - val_loss: 0.6495 - val_acc: 0.9022\n",
      "Epoch 57/60\n",
      "12060/12060 [==============================] - 199s 16ms/step - loss: 0.3665 - acc: 0.9822 - val_loss: 0.6337 - val_acc: 0.9072\n",
      "Epoch 58/60\n",
      "12060/12060 [==============================] - 197s 16ms/step - loss: 0.3628 - acc: 0.9818 - val_loss: 0.6353 - val_acc: 0.9052\n",
      "Epoch 59/60\n",
      "12060/12060 [==============================] - 197s 16ms/step - loss: 0.3585 - acc: 0.9832 - val_loss: 0.6378 - val_acc: 0.9058\n",
      "Epoch 60/60\n",
      "12060/12060 [==============================] - 197s 16ms/step - loss: 0.3603 - acc: 0.9835 - val_loss: 0.6323 - val_acc: 0.9088\n"
     ]
    }
   ],
   "source": [
    "# 模型结构：CNN-LSTM\n",
    "from keras.layers.core import *\n",
    "model = Sequential()\n",
    "model.add(Embedding(len(word_index) + 1,embedding_dim, weights=[embedding_matrix],input_length=MAX_SEQUENCE_LENGTH,))\n",
    "model.add(SpatialDropout1D(0.3))\n",
    "model.add(Conv1D(activation=\"relu\", padding=\"same\", filters=128, kernel_size=5))\n",
    "model.add(MaxPooling1D(pool_size=4))\n",
    "model.add(LSTM(100, dropout=0.3, recurrent_dropout=0.2))\n",
    "model.add(Dense(20, activation='softmax',kernel_regularizer=regularizers.l2(0.01), \n",
    "                activity_regularizer=regularizers.l1(0.001)))\n",
    "model.compile(optimizer=\"adam\", loss='categorical_crossentropy', metrics=['accuracy'])\n",
    "print(model.summary())\n",
    "history_rnn = model.fit(x_train, y_train, batch_size=128, epochs=60,\n",
    "                        validation_data=(x_val, y_val))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x1c5b897b70>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# 绘制训练和验证曲线\n",
    "plt.plot(history_rnn.history['acc'])\n",
    "plt.plot(history_rnn.history['val_acc'])\n",
    "\n",
    "plt.legend(['Train line', 'Val line'], loc='lower right', fontsize=16)\n",
    "\n",
    "plt.title('CNN-LSTM-Dense Correct Rate', fontsize=20)\n",
    "plt.xlabel('Epoch Count', fontsize=18)\n",
    "plt.ylabel('Correct Rate', fontsize=18)\n",
    "\n",
    "plt.xticks(range(0, 61, 5), fontsize=16)\n",
    "plt.yticks(fontsize=16)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "3770/3770 [==============================] - 15s 4ms/step\n",
      "Test score: 0.5894066171083273\n",
      "Test accuracy: 0.9177718833207452\n"
     ]
    }
   ],
   "source": [
    "# 模型测试 评估\n",
    "score, acc = model.evaluate(x_test, y_test, batch_size=128)\n",
    "print('Test score:', score)\n",
    "print('Test accuracy:', acc)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "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": "3.6.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}