Selldetection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Bài toán xác định bài viết bán hàng \n",
    " \n",
    "### Tóm tắt yêu cầu:\n",
    "Xác định một bài viết trên mạng xã hội có phải là bán hàng hay không?\n",
    "### Bộ dữ liệu:\n",
    "60.000 dòng trạng thái trên mạng xã hội đã được gẵn nhãn bán hàng hay không.\n",
    "\n",
    "## Thực hiện\n",
    "\n",
    "### Tiền xử lí dữ liệu\n",
    "\n",
    "Thêm các thư viện cần sử dụng:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from collections import defaultdict\n",
    "from sklearn.feature_extraction.text import TfidfVectorizer\n",
    "from sklearn import model_selection, naive_bayes, svm\n",
    "from sklearn.metrics import accuracy_score\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Đọc dữ liệu đầu vào và tách ra những data bán hàng và không bán hàng\n",
    "Nhìn qua bộ dữ liệu, chuyển các nhãn thành số:\n",
    "* 1 = không bán hàng\n",
    "* 0 = bán hàng"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>Gsu khsi nek dep gai hk cho y kirn yk nhoaz kh...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>Haiz ... chan wa. hoc-hoc-nua-hoc mau</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>Găng tay ngập shop</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>🌺 2 siểu phẩm XÔNG BẮC THẢO DƯỢC 🌺 ( Công dụng...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>Nổi nhớ dân trào 😂😂😂</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   label                                            content\n",
       "0      1  Gsu khsi nek dep gai hk cho y kirn yk nhoaz kh...\n",
       "1      1              Haiz ... chan wa. hoc-hoc-nua-hoc mau\n",
       "2      1                                 Găng tay ngập shop\n",
       "3      0  🌺 2 siểu phẩm XÔNG BẮC THẢO DƯỢC 🌺 ( Công dụng...\n",
       "4      1                               Nổi nhớ dân trào 😂😂😂"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "SELL = '__label__post_ban_hang'\n",
    "NON_SELL = '__label__post_khong_ban'\n",
    "data  = pd.read_csv('./data/data_original.txt', header = None, sep=\"\\n\")\n",
    "tmp = data[0].str.split(\" \", 1, expand = True)\n",
    "df = pd.DataFrame(columns=['label', 'content'])\n",
    "data['label'] = tmp[0]\n",
    "data['content'] = tmp[1]\n",
    "data.drop([data.columns[0]], axis='columns', inplace=True)\n",
    "Encoder = LabelEncoder()\n",
    "data['label'] = Encoder.fit_transform(data['label'])\n",
    "data.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Số lượng dữ liệu và từng loại label:\n",
    "* Bán hàng: 7935 (chiếm 13.2%)\n",
    "* Không bán hàng: 52065 (chiếm 86.8%)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>label</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>7935</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>52065</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       content\n",
       "label         \n",
       "0         7935\n",
       "1        52065"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.groupby('label').count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Xem những từ xuất hiện nhiều trên bộ dữ liệu:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%run ./pre-process/top_words\n",
    "top_words(data['content'].tolist(), file_name = 'top_before_pre')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Làm đẹp dữ liệu text qua các bước:\n",
    "* Chuyển thành kí tự viết thường\n",
    "* Loại bỏ các emoji\n",
    "* Chuyển đổi các định dạng giá cả $\\Rightarrow$ __giá_sản_phẩm__\n",
    "   * Ví dụ: 500k, 500.000đ, 500.000vnđ, 5m, 5tr6, ...\n",
    "* Chuyển đổi link website \t$\\Rightarrow$ __link_website__\n",
    "   * Toàn bộ link bắt đầu từ http, https, bit.ly, goo.gl\n",
    "* Chuyển đổi các dạng số điện thoại \t$\\Rightarrow$ __số_điện_thoại__\n",
    "   * Ví dụ: +84946.123.45, 096-265-222, 096.265.222, ...\n",
    "* Xóa các dấu câu, kí tự lạ.\n",
    "* Xoá kí tự số\n",
    "* Tokenizer\n",
    "   * Chuyển các từ ghép có nghĩa thành một từ: xin chào $\\rightarrow$ __xin_chào__, tóm tắt $\\rightarrow$ __tóm_tắt__, kính đổi màu $\\rightarrow$ __kính_đổi_màu__ ... \n",
    "* Xóa các data NaN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Loading model from file pre-process/vi-segmentation.crfsuite\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>label</th>\n",
       "      <th>content</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>gsu khsi nek dep gai hk cho y kirn yk nhoaz kh...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>haiz chan wa hoc hoc nua hoc mau</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>găng_tay ngập shop</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>siểu phẩm xông bắc thảo_dược công_dụng từng lo...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>1</td>\n",
       "      <td>nổi nhớ dân trào</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59995</td>\n",
       "      <td>1</td>\n",
       "      <td>còn ít mầu nữa</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59996</td>\n",
       "      <td>0</td>\n",
       "      <td>đầm xòe cổ sen sát_nách chất da cá sỉ giá_sản_...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59997</td>\n",
       "      <td>1</td>\n",
       "      <td>người đẹp chuyển giới thái lan biểu_diễn</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59998</td>\n",
       "      <td>1</td>\n",
       "      <td>do not understand why this walking throughout ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59999</td>\n",
       "      <td>1</td>\n",
       "      <td>hòa chung không_khí rộn_ràng của cả nước u23 v...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>60000 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       label                                            content\n",
       "0          1  gsu khsi nek dep gai hk cho y kirn yk nhoaz kh...\n",
       "1          1                   haiz chan wa hoc hoc nua hoc mau\n",
       "2          1                                 găng_tay ngập shop\n",
       "3          0  siểu phẩm xông bắc thảo_dược công_dụng từng lo...\n",
       "4          1                                   nổi nhớ dân trào\n",
       "...      ...                                                ...\n",
       "59995      1                                     còn ít mầu nữa\n",
       "59996      0  đầm xòe cổ sen sát_nách chất da cá sỉ giá_sản_...\n",
       "59997      1           người đẹp chuyển giới thái lan biểu_diễn\n",
       "59998      1  do not understand why this walking throughout ...\n",
       "59999      1  hòa chung không_khí rộn_ràng của cả nước u23 v...\n",
       "\n",
       "[60000 rows x 2 columns]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%run ./pre-process/pre_process\n",
    "data['content'] = data['content'].map(lambda s : pre_process(s))\n",
    "data.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "10    sữa rửa mặt ohui miracle moisture sữa rửa mặt ...\n",
       "11    chuc mừng sinh_nhật chj le nguyen em tuong vy ...\n",
       "12    clutch đà_điểu sang_trọng cho nam call số_điện...\n",
       "13    quần_jean chân viền ren hoa giá_sản_phẩm chất ...\n",
       "14               mới_đó mà đã chủ_nhật cuối tuần vui_vẻ\n",
       "15                                    nkìn a từ fía sau\n",
       "16                                       chuc thu sn vv\n",
       "17                         quả set up xấu thê_thảm nhất\n",
       "18                         vẫn người chụp ảnh có tâm ấy\n",
       "19    đầm ren hoa sang_trọng và tinh_tế và thanh_lịc...\n",
       "20    kam on may pn to chuc sn mk tuy muon nhưng mk ...\n",
       "21    một ngày lễ thật vui và ý_nghĩa chúc tất_cả th...\n",
       "22    hom nay la ngay vui ve nhat cua gia dinh da co...\n",
       "23                 hay tu nhien va vo tu nhu vay tot ko\n",
       "24    tàn_nhang bẩm_sinh đã hết nhờ thảo_dược hồng h...\n",
       "25    bộ đồ_chơi xúc xắc giá giá_sản_phẩm có số_lượn...\n",
       "26                                                     \n",
       "27    code ngày mã code2103 lông_vũ locke thần vũ lo...\n",
       "28    e lam co vui ko o no do co nguoi van mong cho ...\n",
       "29                                       ghi gì thì ghi\n",
       "Name: content, dtype: object"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data['content'][10:30]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Sau khi đã xử lí dữ liệu text, xem những từ xuất hiện nhiều nhất trong bộ dữ liệu:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%run ./pre-process/top_words\n",
    "two_top_words(data, file_name = 'top_after_pre')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Khi nhìn vào biểu đồ những từ xuất hiện nhiều trong dữ liệu, ta dễ dàng thấy một số từ không mang tính phân loại hoặc từ không mang nhiều ý nghĩa, ví dụ: *và*, *là*, *có*, *cho*, ...\n",
    "\n",
    "Tiến hành loại bỏ những từ này bằng cách sử dụng bộ từ điển *stopwords*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "%run ./pre-process/top_words\n",
    "\n",
    "stopwords = set(line.strip() for line in open('pre-process/stopwords.txt'))\n",
    "data['content'] = data['content'].apply(lambda x: ' '.join([word for word in x.split() if word not in (stopwords)]))\n",
    "two_top_words(data, file_name = 'top_after_pre')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Vector hóa, train và đánh giá phương pháp:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# Tách dữ liệu thành tập train và test\n",
    "Train_X, Test_X, Train_Y, Test_Y = model_selection.train_test_split(data['content'], data['label'],test_size=0.3, random_state=42)\n",
    "\n",
    "# Vecto hóa dùng TF-IDF Vectorizer\n",
    "Tfidf_vect = TfidfVectorizer(max_features=5000, lowercase=False)\n",
    "Tfidf_vect.fit(data['content'])\n",
    "\n",
    "Train_X_Tfidf = Tfidf_vect.transform(Train_X)\n",
    "Test_X_Tfidf = Tfidf_vect.transform(Test_X)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Naive Bayes Accuracy ->  96.83333333333334\n",
      "SVM Accuracy ->  97.71111111111111\n"
     ]
    }
   ],
   "source": [
    "# Phân loại dùng Naive Bayes và SVM để so sánh\n",
    "\n",
    "# Naive Bayes\n",
    "# fit tập train \n",
    "Naive = naive_bayes.MultinomialNB()\n",
    "Naive.fit(Train_X_Tfidf,Train_Y)\n",
    "\n",
    "# dự đoán\n",
    "predictions_NB = Naive.predict(Test_X_Tfidf)\n",
    "\n",
    "#  Tính Accuracy\n",
    "print(\"Naive Bayes Accuracy -> \",accuracy_score(predictions_NB, Test_Y)*100)\n",
    "\n",
    "\n",
    "# SVM\n",
    "# fit tập train \n",
    "SVM = svm.SVC(C=1.0, kernel='linear', degree=3, gamma='auto')\n",
    "SVM.fit(Train_X_Tfidf,Train_Y)\n",
    "\n",
    "# dự đoán\n",
    "predictions_SVM = SVM.predict(Test_X_Tfidf)\n",
    "\n",
    "# Tính Accuracy\n",
    "print(\"SVM Accuracy -> \",accuracy_score(predictions_SVM, Test_Y)*100)"
   ]
  }
 ],
 "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": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}