02_entity_resolution_no_solution.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# End-to-End Entity Resolution with pyJedAI and Magellan"
   ]
  },
  {
   "attachments": {
    "image.png": {
     "image/png": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![image.png](attachment:image.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load the dataset and the ground truth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import warnings\n",
    "warnings.simplefilter(action='ignore', category=FutureWarning)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>name</th>\n",
       "      <th>brand</th>\n",
       "      <th>size</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>Lexar 32GB 1400x 210MB/s professional XQD memÃ...</td>\n",
       "      <td>LEXAR</td>\n",
       "      <td>32.0</td>\n",
       "      <td>31990.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_001</td>\n",
       "      <td>Sony microSDXC UHS-1 U3 128GB memÃ³riakÃ¡rtya ...</td>\n",
       "      <td>SONY</td>\n",
       "      <td>128.0</td>\n",
       "      <td>59990.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_002</td>\n",
       "      <td>SANDISK DUAL DRIVE, TYPE-C, USB 3.0, 16GB, 130...</td>\n",
       "      <td>SANDISK</td>\n",
       "      <td>16.0</td>\n",
       "      <td>6790.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_003</td>\n",
       "      <td>SanDisk Dual Drive, TYPE-C, USB 3.0, 64GB, 150...</td>\n",
       "      <td>SANDISK</td>\n",
       "      <td>64.0</td>\n",
       "      <td>14190.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_004</td>\n",
       "      <td>Lexar XQD 32GB X1400 PROFESSIONAL XQD +KUPON P...</td>\n",
       "      <td>LEXAR</td>\n",
       "      <td>32.0</td>\n",
       "      <td>329.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>830</th>\n",
       "      <td>altosight_830</td>\n",
       "      <td>Pendrive LEXAR Jumpdrive S25 64 GB</td>\n",
       "      <td>LEXAR</td>\n",
       "      <td>64.0</td>\n",
       "      <td>24.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>831</th>\n",
       "      <td>altosight_831</td>\n",
       "      <td>Pendrive SANDISK Extreme GO 3.1 256GB</td>\n",
       "      <td>SANDISK</td>\n",
       "      <td>256.0</td>\n",
       "      <td>195.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>832</th>\n",
       "      <td>altosight_832</td>\n",
       "      <td>Tarjeta de Memoria SD  SONY 16GB EXPERIENCE</td>\n",
       "      <td>SONY</td>\n",
       "      <td>16.0</td>\n",
       "      <td>15.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>833</th>\n",
       "      <td>altosight_833</td>\n",
       "      <td>Tarjeta de Memoria SANDISK Extreme Pro SDHC 8GB</td>\n",
       "      <td>SANDISK</td>\n",
       "      <td>8.0</td>\n",
       "      <td>22.97</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>834</th>\n",
       "      <td>altosight_834</td>\n",
       "      <td>Tarjeta de Memoria SANDISK SDHC 8GB</td>\n",
       "      <td>SANDISK</td>\n",
       "      <td>8.0</td>\n",
       "      <td>4.99</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>835 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               _id                                               name  \\\n",
       "0    altosight_000  Lexar 32GB 1400x 210MB/s professional XQD memÃ...   \n",
       "1    altosight_001  Sony microSDXC UHS-1 U3 128GB memÃ³riakÃ¡rtya ...   \n",
       "2    altosight_002  SANDISK DUAL DRIVE, TYPE-C, USB 3.0, 16GB, 130...   \n",
       "3    altosight_003  SanDisk Dual Drive, TYPE-C, USB 3.0, 64GB, 150...   \n",
       "4    altosight_004  Lexar XQD 32GB X1400 PROFESSIONAL XQD +KUPON P...   \n",
       "..             ...                                                ...   \n",
       "830  altosight_830                 Pendrive LEXAR Jumpdrive S25 64 GB   \n",
       "831  altosight_831              Pendrive SANDISK Extreme GO 3.1 256GB   \n",
       "832  altosight_832        Tarjeta de Memoria SD  SONY 16GB EXPERIENCE   \n",
       "833  altosight_833    Tarjeta de Memoria SANDISK Extreme Pro SDHC 8GB   \n",
       "834  altosight_834                Tarjeta de Memoria SANDISK SDHC 8GB   \n",
       "\n",
       "       brand   size     price  \n",
       "0      LEXAR   32.0  31990.00  \n",
       "1       SONY  128.0  59990.00  \n",
       "2    SANDISK   16.0   6790.00  \n",
       "3    SANDISK   64.0  14190.00  \n",
       "4      LEXAR   32.0    329.00  \n",
       "..       ...    ...       ...  \n",
       "830    LEXAR   64.0     24.99  \n",
       "831  SANDISK  256.0    195.99  \n",
       "832     SONY   16.0     15.99  \n",
       "833  SANDISK    8.0     22.97  \n",
       "834  SANDISK    8.0      4.99  \n",
       "\n",
       "[835 rows x 5 columns]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path_data = \"datasets/altosight_usb_sticks_small/\"\n",
    "ds = pd.read_csv(path_data + \"dataset.csv\")\n",
    "ds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>name</th>\n",
       "      <th>brand</th>\n",
       "      <th>size</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>lexar 32gb 1400x 210mb/s professional xqd memã...</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>31990.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_001</td>\n",
       "      <td>sony microsdxc uhs-1 u3 128gb memã³riakã¡rtya ...</td>\n",
       "      <td>sony</td>\n",
       "      <td>128.0</td>\n",
       "      <td>59990.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_002</td>\n",
       "      <td>sandisk dual drive, type-c, usb 3.0, 16gb, 130...</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>16.0</td>\n",
       "      <td>6790.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_003</td>\n",
       "      <td>sandisk dual drive, type-c, usb 3.0, 64gb, 150...</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>64.0</td>\n",
       "      <td>14190.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_004</td>\n",
       "      <td>lexar xqd 32gb x1400 professional xqd +kupon p...</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>329.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>830</th>\n",
       "      <td>altosight_830</td>\n",
       "      <td>pendrive lexar jumpdrive s25 64 gb</td>\n",
       "      <td>lexar</td>\n",
       "      <td>64.0</td>\n",
       "      <td>24.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>831</th>\n",
       "      <td>altosight_831</td>\n",
       "      <td>pendrive sandisk extreme go 3.1 256gb</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>256.0</td>\n",
       "      <td>195.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>832</th>\n",
       "      <td>altosight_832</td>\n",
       "      <td>tarjeta de memoria sd  sony 16gb experience</td>\n",
       "      <td>sony</td>\n",
       "      <td>16.0</td>\n",
       "      <td>15.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>833</th>\n",
       "      <td>altosight_833</td>\n",
       "      <td>tarjeta de memoria sandisk extreme pro sdhc 8gb</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>8.0</td>\n",
       "      <td>22.97</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>834</th>\n",
       "      <td>altosight_834</td>\n",
       "      <td>tarjeta de memoria sandisk sdhc 8gb</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>8.0</td>\n",
       "      <td>4.99</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>835 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               _id                                               name  \\\n",
       "0    altosight_000  lexar 32gb 1400x 210mb/s professional xqd memã...   \n",
       "1    altosight_001  sony microsdxc uhs-1 u3 128gb memã³riakã¡rtya ...   \n",
       "2    altosight_002  sandisk dual drive, type-c, usb 3.0, 16gb, 130...   \n",
       "3    altosight_003  sandisk dual drive, type-c, usb 3.0, 64gb, 150...   \n",
       "4    altosight_004  lexar xqd 32gb x1400 professional xqd +kupon p...   \n",
       "..             ...                                                ...   \n",
       "830  altosight_830                 pendrive lexar jumpdrive s25 64 gb   \n",
       "831  altosight_831              pendrive sandisk extreme go 3.1 256gb   \n",
       "832  altosight_832        tarjeta de memoria sd  sony 16gb experience   \n",
       "833  altosight_833    tarjeta de memoria sandisk extreme pro sdhc 8gb   \n",
       "834  altosight_834                tarjeta de memoria sandisk sdhc 8gb   \n",
       "\n",
       "       brand   size     price  \n",
       "0      lexar   32.0  31990.00  \n",
       "1       sony  128.0  59990.00  \n",
       "2    sandisk   16.0   6790.00  \n",
       "3    sandisk   64.0  14190.00  \n",
       "4      lexar   32.0    329.00  \n",
       "..       ...    ...       ...  \n",
       "830    lexar   64.0     24.99  \n",
       "831  sandisk  256.0    195.99  \n",
       "832     sony   16.0     15.99  \n",
       "833  sandisk    8.0     22.97  \n",
       "834  sandisk    8.0      4.99  \n",
       "\n",
       "[835 rows x 5 columns]"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "str_attrs = [\"name\", \"brand\"]\n",
    "for attr in str_attrs:\n",
    "    ds[attr] = ds[attr].apply(lambda x: x.lower() if isinstance(x, str) else x)\n",
    "ds"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "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>l_id</th>\n",
       "      <th>r_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>altosight_004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>altosight_005</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>altosight_013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>altosight_047</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>altosight_161</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4077</th>\n",
       "      <td>altosight_806</td>\n",
       "      <td>altosight_821</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4078</th>\n",
       "      <td>altosight_806</td>\n",
       "      <td>altosight_832</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4079</th>\n",
       "      <td>altosight_809</td>\n",
       "      <td>altosight_825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4080</th>\n",
       "      <td>altosight_817</td>\n",
       "      <td>altosight_823</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4081</th>\n",
       "      <td>altosight_821</td>\n",
       "      <td>altosight_832</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>4082 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "               l_id           r_id\n",
       "0     altosight_000  altosight_004\n",
       "1     altosight_000  altosight_005\n",
       "2     altosight_000  altosight_013\n",
       "3     altosight_000  altosight_047\n",
       "4     altosight_000  altosight_161\n",
       "...             ...            ...\n",
       "4077  altosight_806  altosight_821\n",
       "4078  altosight_806  altosight_832\n",
       "4079  altosight_809  altosight_825\n",
       "4080  altosight_817  altosight_823\n",
       "4081  altosight_821  altosight_832\n",
       "\n",
       "[4082 rows x 2 columns]"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gold = pd.read_csv(path_data + \"matches.csv\")\n",
    "gold"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>name</th>\n",
       "      <th>brand</th>\n",
       "      <th>size</th>\n",
       "      <th>price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_000</td>\n",
       "      <td>lexar 32gb 1400x 210mb/s professional xqd memã...</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>31990.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_004</td>\n",
       "      <td>lexar xqd 32gb x1400 professional xqd +kupon p...</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>329.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             _id                                               name  brand  \\\n",
       "0  altosight_000  lexar 32gb 1400x 210mb/s professional xqd memã...  lexar   \n",
       "4  altosight_004  lexar xqd 32gb x1400 professional xqd +kupon p...  lexar   \n",
       "\n",
       "   size    price  \n",
       "0  32.0  31990.0  \n",
       "4  32.0    329.0  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ds[(ds[\"_id\"] == \"altosight_000\") | (ds[\"_id\"] == \"altosight_004\")]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Blocking (pyJedAI)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "attrs = [\"_id\", \"name\", \"brand\", \"size\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.datamodel import Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = Data(\n",
    "    dataset_1=ds,\n",
    "    id_column_name_1=\"_id\",\n",
    "    ground_truth=gold,\n",
    "    attributes_1=attrs\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Block Building"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.block_building import (\n",
    "    StandardBlocking,\n",
    "    QGramsBlocking,\n",
    "    ExtendedQGramsBlocking,\n",
    "    SuffixArraysBlocking,\n",
    "    ExtendedSuffixArraysBlocking\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5dc49d097aad440d99a7132c243367ce",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Standard Blocking:   0%|          | 0/835 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "bb = StandardBlocking()\n",
    "blocks = bb.build_blocks(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***************************************************************************************************************************\n",
      "                                         Μethod:  Standard Blocking\n",
      "***************************************************************************************************************************\n",
      "Method name: Standard Blocking\n",
      "Parameters: \n",
      "Runtime: 0.2624 seconds\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n",
      "Performance:\n",
      "\tPrecision:      0.38% \n",
      "\tRecall:       100.00%\n",
      "\tF1-score:       0.75%\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n"
     ]
    }
   ],
   "source": [
    "_ = bb.evaluate(blocks)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Block Purging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.block_cleaning import BlockPurging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "85471338ec7a491d8b860184568045ec",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Block Purging:   0%|          | 0/1228 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "bp = BlockPurging(smoothing_factor=1.05)\n",
    "purging_blocks = bp.process(blocks, data, tqdm_disable=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***************************************************************************************************************************\n",
      "                                         Μethod:  Block Purging\n",
      "***************************************************************************************************************************\n",
      "Method name: Block Purging\n",
      "Parameters: \n",
      "\tSmoothing factor: 1.05\n",
      "\tMax Comparisons per Block: 16653.0\n",
      "Runtime: 0.0510 seconds\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n",
      "Performance:\n",
      "\tPrecision:      1.34% \n",
      "\tRecall:       100.00%\n",
      "\tF1-score:       2.64%\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n"
     ]
    }
   ],
   "source": [
    "_ = bp.evaluate(purging_blocks)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Block Filtering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.block_cleaning import BlockFiltering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5b999d048cad46609846e4ed5795e1fb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Block Filtering:   0%|          | 0/3 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "bf = BlockFiltering(ratio=0.95)\n",
    "filtering_blocks = bf.process(purging_blocks, data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***************************************************************************************************************************\n",
      "                                         Μethod:  Block Filtering\n",
      "***************************************************************************************************************************\n",
      "Method name: Block Filtering\n",
      "Parameters: \n",
      "\tRatio: 0.95\n",
      "Runtime: 0.0828 seconds\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n",
      "Performance:\n",
      "\tPrecision:      1.58% \n",
      "\tRecall:        99.76%\n",
      "\tF1-score:       3.12%\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n"
     ]
    }
   ],
   "source": [
    "_ = bf.evaluate(filtering_blocks)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Comparison Cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.comparison_cleaning import (\n",
    "    ComparisonPropagation,\n",
    "    CardinalityEdgePruning,\n",
    "    CardinalityNodePruning,\n",
    "    WeightedEdgePruning,\n",
    "    WeightedNodePruning,\n",
    "    ReciprocalCardinalityNodePruning,\n",
    "    ReciprocalWeightedNodePruning,\n",
    "    BLAST\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3af1019022b0401ebe5fae0689ba710f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Weighted Edge Pruning:   0%|          | 0/835 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "mb = WeightedEdgePruning(weighting_scheme='CBS')\n",
    "comparisons = mb.process(filtering_blocks, data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***************************************************************************************************************************\n",
      "                                         Μethod:  Weighted Edge Pruning\n",
      "***************************************************************************************************************************\n",
      "Method name: Weighted Edge Pruning\n",
      "Parameters: \n",
      "\tNode centric: False\n",
      "\tWeighting scheme: CBS\n",
      "Runtime: 1.0664 seconds\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n",
      "Performance:\n",
      "\tPrecision:      6.49% \n",
      "\tRecall:        97.94%\n",
      "\tF1-score:      12.18%\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n"
     ]
    }
   ],
   "source": [
    "_ = mb.evaluate(comparisons)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generate the candidate set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'altosight_000'"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "id_mapping = {d[1]: d[0] for d in data._ids_mapping_1.items()}\n",
    "id_mapping[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "import random"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('altosight_129', 'altosight_516')\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "61564"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "candidates = {(id_mapping[cand[0]], id_mapping[cand[1]]) if id_mapping[cand[0]] < id_mapping[cand[1]]\n",
    "              else (id_mapping[cand[1]], id_mapping[cand[0]]) for cand in [(l_id, r_id) for r_id in comparisons.keys() for l_id in comparisons[r_id]]}\n",
    "print(random.choice(list(candidates)))\n",
    "len(candidates)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "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>l_id</th>\n",
       "      <th>r_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_114</td>\n",
       "      <td>altosight_702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_513</td>\n",
       "      <td>altosight_563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_142</td>\n",
       "      <td>altosight_613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_085</td>\n",
       "      <td>altosight_091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_297</td>\n",
       "      <td>altosight_718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61559</th>\n",
       "      <td>altosight_173</td>\n",
       "      <td>altosight_743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61560</th>\n",
       "      <td>altosight_199</td>\n",
       "      <td>altosight_707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61561</th>\n",
       "      <td>altosight_049</td>\n",
       "      <td>altosight_569</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61562</th>\n",
       "      <td>altosight_204</td>\n",
       "      <td>altosight_684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61563</th>\n",
       "      <td>altosight_089</td>\n",
       "      <td>altosight_657</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61564 rows × 2 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                l_id           r_id\n",
       "0      altosight_114  altosight_702\n",
       "1      altosight_513  altosight_563\n",
       "2      altosight_142  altosight_613\n",
       "3      altosight_085  altosight_091\n",
       "4      altosight_297  altosight_718\n",
       "...              ...            ...\n",
       "61559  altosight_173  altosight_743\n",
       "61560  altosight_199  altosight_707\n",
       "61561  altosight_049  altosight_569\n",
       "61562  altosight_204  altosight_684\n",
       "61563  altosight_089  altosight_657\n",
       "\n",
       "[61564 rows x 2 columns]"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "candidates = pd.DataFrame(candidates, columns=[\"l_id\", \"r_id\"])\n",
    "candidates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "candidates.to_csv(\"candidates.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Case without comparison cleaning\n",
    "\n",
    "# import itertools\n",
    "# import random\n",
    "\n",
    "# candidates = set()\n",
    "# for key in list(filtering_blocks.keys()):\n",
    "#     pairs = itertools.combinations(list(filtering_blocks[key].entities_D1), 2)\n",
    "#     for pair in pairs:\n",
    "#         candidates.add((id_mapping[pair[0]], id_mapping[pair[1]]) if id_mapping[pair[0]] < id_mapping[pair[1]]\n",
    "#                        else (id_mapping[pair[1]], id_mapping[pair[0]]))\n",
    "# print(random.choice(list(candidates)))\n",
    "# len(candidates)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Entity Matching (Magellan)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "import py_entitymatching as em"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Prepare the input dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Specify numeric columns and cast any other value to string."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "_id       object\n",
       "name      object\n",
       "brand     object\n",
       "size     float64\n",
       "price    float64\n",
       "dtype: object"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "num_cols = [\"size\", \"price\"]\n",
    "for col in num_cols:\n",
    "    ds[col] = pd.to_numeric(ds[col])\n",
    "for col in ds.columns:\n",
    "    if ds[col].dtype == \"object\":\n",
    "        ds[col] = ds[col].str.lower()\n",
    "ds.dtypes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set the key and collect useful attributes (excluding the ID)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['name', 'brand', 'size', 'price']"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "em.set_key(ds, \"_id\")\n",
    "attrs = [attr for attr in ds.columns if attr != \"_id\"]\n",
    "attrs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load and prepare the labeled dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>altosight_134</td>\n",
       "      <td>altosight_273</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>altosight_424</td>\n",
       "      <td>altosight_664</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>altosight_322</td>\n",
       "      <td>altosight_542</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>altosight_558</td>\n",
       "      <td>altosight_677</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>altosight_267</td>\n",
       "      <td>altosight_712</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8542</th>\n",
       "      <td>8542</td>\n",
       "      <td>altosight_576</td>\n",
       "      <td>altosight_809</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8543</th>\n",
       "      <td>8543</td>\n",
       "      <td>altosight_069</td>\n",
       "      <td>altosight_796</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8544</th>\n",
       "      <td>8544</td>\n",
       "      <td>altosight_612</td>\n",
       "      <td>altosight_679</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8545</th>\n",
       "      <td>8545</td>\n",
       "      <td>altosight_172</td>\n",
       "      <td>altosight_666</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8546</th>\n",
       "      <td>8546</td>\n",
       "      <td>altosight_445</td>\n",
       "      <td>altosight_760</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8547 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       _id      ltable_id      rtable_id  label\n",
       "0        0  altosight_134  altosight_273      1\n",
       "1        1  altosight_424  altosight_664      1\n",
       "2        2  altosight_322  altosight_542      1\n",
       "3        3  altosight_558  altosight_677      1\n",
       "4        4  altosight_267  altosight_712      1\n",
       "...    ...            ...            ...    ...\n",
       "8542  8542  altosight_576  altosight_809      0\n",
       "8543  8543  altosight_069  altosight_796      0\n",
       "8544  8544  altosight_612  altosight_679      0\n",
       "8545  8545  altosight_172  altosight_666      0\n",
       "8546  8546  altosight_445  altosight_760      0\n",
       "\n",
       "[8547 rows x 4 columns]"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels = pd.read_csv(path_data + \"labels.csv\")\n",
    "labels = labels.reset_index()\n",
    "labels = labels.rename(columns={\"index\": \"_id\", \"l_id\": \"ltable_id\", \"r_id\": \"rtable_id\"})\n",
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>label</th>\n",
       "      <th>ltable_name</th>\n",
       "      <th>ltable_brand</th>\n",
       "      <th>ltable_size</th>\n",
       "      <th>ltable_price</th>\n",
       "      <th>rtable_name</th>\n",
       "      <th>rtable_brand</th>\n",
       "      <th>rtable_size</th>\n",
       "      <th>rtable_price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>altosight_134</td>\n",
       "      <td>altosight_273</td>\n",
       "      <td>1</td>\n",
       "      <td>intenso 3503470 basic line 16 gb usb-stick 16 gb</td>\n",
       "      <td>intenso</td>\n",
       "      <td>16.0</td>\n",
       "      <td>7.99</td>\n",
       "      <td>intenso basic line 3503470 16gb usb 2.0 flash drive</td>\n",
       "      <td>intenso</td>\n",
       "      <td>16.0</td>\n",
       "      <td>8.03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>altosight_424</td>\n",
       "      <td>altosight_664</td>\n",
       "      <td>1</td>\n",
       "      <td>carte mã©moire microsdhc sandisk ultra plus 128 go, classe 10, vitesse de lecture allant jusqu'ã...</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>128.0</td>\n",
       "      <td>74.99</td>\n",
       "      <td>carte mã©moire sandisk ultra plus microsdxc 128go 80mo/seconde uhs-i avec adaptateur sd\\n</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>128.0</td>\n",
       "      <td>80.04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>altosight_322</td>\n",
       "      <td>altosight_542</td>\n",
       "      <td>1</td>\n",
       "      <td>lexar ljdc20m-128bbeuk jumpdrive c20m micro-usb-flash-laufwerk 128gb</td>\n",
       "      <td>lexar</td>\n",
       "      <td>128.0</td>\n",
       "      <td>53.61</td>\n",
       "      <td>clã© usb lexar jumpdrive c20m 128gb 128gb usb 3.0 (3.1 gen 1) type-a noir, argent lecteur usb fl...</td>\n",
       "      <td>lexar</td>\n",
       "      <td>128.0</td>\n",
       "      <td>57.07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>altosight_558</td>\n",
       "      <td>altosight_677</td>\n",
       "      <td>1</td>\n",
       "      <td>intenso usb-stick 3.0 premium line 64gb usb 3.0</td>\n",
       "      <td>intenso</td>\n",
       "      <td>64.0</td>\n",
       "      <td>34.99</td>\n",
       "      <td>intenso premium line 64gb usb 3.0 (3534490)\\n</td>\n",
       "      <td>intenso</td>\n",
       "      <td>64.0</td>\n",
       "      <td>41.15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>altosight_267</td>\n",
       "      <td>altosight_712</td>\n",
       "      <td>1</td>\n",
       "      <td>toshiba exceria pro n401 64gb sd memory card 95 mb/s 4k hd - thn-n401s0640e4</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>64.0</td>\n",
       "      <td>41.47</td>\n",
       "      <td>toshiba sdhc exceria pro 64gb, silber</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>64.0</td>\n",
       "      <td>59.95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8542</th>\n",
       "      <td>8542</td>\n",
       "      <td>altosight_576</td>\n",
       "      <td>altosight_809</td>\n",
       "      <td>0</td>\n",
       "      <td>toshiba exceria n302 sdxc kort 128 gb - minneskort till kamera</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>128.0</td>\n",
       "      <td>699.00</td>\n",
       "      <td>tesco direct: tos-u202-128gb-whi 128gb transmemory usb 2.0 flash drive</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>128.0</td>\n",
       "      <td>34.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8543</th>\n",
       "      <td>8543</td>\n",
       "      <td>altosight_069</td>\n",
       "      <td>altosight_796</td>\n",
       "      <td>0</td>\n",
       "      <td>sandisk - cruzer extreme usb 3.0 16gb&amp;nbsp;</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>16.0</td>\n",
       "      <td>29.90</td>\n",
       "      <td>tesco direct: lexar jumpdrive v30 16 gb usb 2.0 flash drive - teal</td>\n",
       "      <td>lexar</td>\n",
       "      <td>16.0</td>\n",
       "      <td>10.13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8544</th>\n",
       "      <td>8544</td>\n",
       "      <td>altosight_612</td>\n",
       "      <td>altosight_679</td>\n",
       "      <td>0</td>\n",
       "      <td>sandisk - cruzer fit 128gb -</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>128.0</td>\n",
       "      <td>129.00</td>\n",
       "      <td>sony compact slim hd-sl1 - disque dur - 1 to - usb 3.0\\n</td>\n",
       "      <td>sony</td>\n",
       "      <td>1024.0</td>\n",
       "      <td>109.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8545</th>\n",
       "      <td>8545</td>\n",
       "      <td>altosight_172</td>\n",
       "      <td>altosight_666</td>\n",
       "      <td>0</td>\n",
       "      <td>samsung galaxy a10 dual sim - 32gb, 2gb ram, 4g lte, black</td>\n",
       "      <td>samsung</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1999.00</td>\n",
       "      <td>carte mã©moire sandisk ultra plus sdhc 16go 80mo/seconde uhs-i\\n</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>16.0</td>\n",
       "      <td>19.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8546</th>\n",
       "      <td>8546</td>\n",
       "      <td>altosight_445</td>\n",
       "      <td>altosight_760</td>\n",
       "      <td>0</td>\n",
       "      <td>toshiba scheda di memoria microsdhc, 32gb - exceria - 95mb/s - uhs-i - u3</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>32.0</td>\n",
       "      <td>31.49</td>\n",
       "      <td>jumpdrive p20 - usb 3.0 - 32 go lexar achat dans le rayon clã©s usb usb 3.0</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>71.78</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>8547 rows × 12 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "       _id      ltable_id      rtable_id  label  \\\n",
       "0        0  altosight_134  altosight_273      1   \n",
       "1        1  altosight_424  altosight_664      1   \n",
       "2        2  altosight_322  altosight_542      1   \n",
       "3        3  altosight_558  altosight_677      1   \n",
       "4        4  altosight_267  altosight_712      1   \n",
       "...    ...            ...            ...    ...   \n",
       "8542  8542  altosight_576  altosight_809      0   \n",
       "8543  8543  altosight_069  altosight_796      0   \n",
       "8544  8544  altosight_612  altosight_679      0   \n",
       "8545  8545  altosight_172  altosight_666      0   \n",
       "8546  8546  altosight_445  altosight_760      0   \n",
       "\n",
       "                                                                                              ltable_name  \\\n",
       "0                                                        intenso 3503470 basic line 16 gb usb-stick 16 gb   \n",
       "1     carte mã©moire microsdhc sandisk ultra plus 128 go, classe 10, vitesse de lecture allant jusqu'ã...   \n",
       "2                                    lexar ljdc20m-128bbeuk jumpdrive c20m micro-usb-flash-laufwerk 128gb   \n",
       "3                                                         intenso usb-stick 3.0 premium line 64gb usb 3.0   \n",
       "4                            toshiba exceria pro n401 64gb sd memory card 95 mb/s 4k hd - thn-n401s0640e4   \n",
       "...                                                                                                   ...   \n",
       "8542                                       toshiba exceria n302 sdxc kort 128 gb - minneskort till kamera   \n",
       "8543                                                          sandisk - cruzer extreme usb 3.0 16gb&nbsp;   \n",
       "8544                                                                         sandisk - cruzer fit 128gb -   \n",
       "8545                                           samsung galaxy a10 dual sim - 32gb, 2gb ram, 4g lte, black   \n",
       "8546                            toshiba scheda di memoria microsdhc, 32gb - exceria - 95mb/s - uhs-i - u3   \n",
       "\n",
       "     ltable_brand  ltable_size  ltable_price  \\\n",
       "0         intenso         16.0          7.99   \n",
       "1         sandisk        128.0         74.99   \n",
       "2           lexar        128.0         53.61   \n",
       "3         intenso         64.0         34.99   \n",
       "4         toshiba         64.0         41.47   \n",
       "...           ...          ...           ...   \n",
       "8542      toshiba        128.0        699.00   \n",
       "8543      sandisk         16.0         29.90   \n",
       "8544      sandisk        128.0        129.00   \n",
       "8545      samsung          NaN       1999.00   \n",
       "8546      toshiba         32.0         31.49   \n",
       "\n",
       "                                                                                              rtable_name  \\\n",
       "0                                                     intenso basic line 3503470 16gb usb 2.0 flash drive   \n",
       "1               carte mã©moire sandisk ultra plus microsdxc 128go 80mo/seconde uhs-i avec adaptateur sd\\n   \n",
       "2     clã© usb lexar jumpdrive c20m 128gb 128gb usb 3.0 (3.1 gen 1) type-a noir, argent lecteur usb fl...   \n",
       "3                                                           intenso premium line 64gb usb 3.0 (3534490)\\n   \n",
       "4                                                                   toshiba sdhc exceria pro 64gb, silber   \n",
       "...                                                                                                   ...   \n",
       "8542                               tesco direct: tos-u202-128gb-whi 128gb transmemory usb 2.0 flash drive   \n",
       "8543                                   tesco direct: lexar jumpdrive v30 16 gb usb 2.0 flash drive - teal   \n",
       "8544                                             sony compact slim hd-sl1 - disque dur - 1 to - usb 3.0\\n   \n",
       "8545                                     carte mã©moire sandisk ultra plus sdhc 16go 80mo/seconde uhs-i\\n   \n",
       "8546                          jumpdrive p20 - usb 3.0 - 32 go lexar achat dans le rayon clã©s usb usb 3.0   \n",
       "\n",
       "     rtable_brand  rtable_size  rtable_price  \n",
       "0         intenso         16.0          8.03  \n",
       "1         sandisk        128.0         80.04  \n",
       "2           lexar        128.0         57.07  \n",
       "3         intenso         64.0         41.15  \n",
       "4         toshiba         64.0         59.95  \n",
       "...           ...          ...           ...  \n",
       "8542      toshiba        128.0         34.99  \n",
       "8543        lexar         16.0         10.13  \n",
       "8544         sony       1024.0        109.00  \n",
       "8545      sandisk         16.0         19.99  \n",
       "8546        lexar         32.0         71.78  \n",
       "\n",
       "[8547 rows x 12 columns]"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "labels = labels.join(ds.set_index(\"_id\"), on=\"ltable_id\")\n",
    "labels = labels.rename(columns={attr: \"ltable_\" + attr for attr in attrs})\n",
    "labels = labels.join(ds.set_index(\"_id\"), on=\"rtable_id\")\n",
    "labels = labels.rename(columns={attr: \"rtable_\" + attr for attr in attrs})\n",
    "em.set_key(labels, \"_id\")\n",
    "labels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels.to_csv(\"labels_expanded.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Split the labeled data into train and test sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Metadata file is not present in the given path; proceeding to read the csv file.\n"
     ]
    }
   ],
   "source": [
    "labels = em.read_csv_metadata(\n",
    "    \"labels_expanded.csv\",\n",
    "    key=\"_id\",\n",
    "    ltable=ds, rtable=ds,\n",
    "    fk_ltable=\"ltable_id\", fk_rtable=\"rtable_id\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_test = em.split_train_test(labels, train_proportion=0.7, random_state=0)\n",
    "train = train_test[\"train\"]\n",
    "test = train_test[\"test\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "ML models supported by Magellan:\n",
    "\n",
    "- Decision Tree: em.DTMatcher(name='DecisionTree', random_state=0)\n",
    "- SVM: em.SVMMatcher(name='SVM', random_state=0)\n",
    "- Random Forest: em.RFMatcher(name='RF', random_state=0)\n",
    "- Logistic Regression: em.LogRegMatcher(name='LogReg', random_state=0)\n",
    "- Linear Regression: em.LinRegMatcher(name='LinReg')\n",
    "- Naive Bayes: em.NBMatcher(name='NaiveBayes')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = em.DTMatcher()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Define the features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['name_name_jac_qgm_3_qgm_3',\n",
       " 'name_name_cos_dlm_dc0_dlm_dc0',\n",
       " 'brand_brand_lev_dist',\n",
       " 'brand_brand_lev_sim',\n",
       " 'brand_brand_jar',\n",
       " 'brand_brand_jwn',\n",
       " 'brand_brand_exm',\n",
       " 'brand_brand_jac_qgm_3_qgm_3',\n",
       " 'size_size_exm',\n",
       " 'size_size_anm',\n",
       " 'size_size_lev_dist',\n",
       " 'size_size_lev_sim',\n",
       " 'price_price_exm',\n",
       " 'price_price_anm',\n",
       " 'price_price_lev_dist',\n",
       " 'price_price_lev_sim']"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "feature_attrs = [\"name\", \"brand\", \"size\", \"price\"]\n",
    "feature_table = em.get_features_for_matching(ds[feature_attrs], ds[feature_attrs], validate_inferred_attr_types=False)\n",
    "list(feature_table[\"feature_name\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate the feature vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['ltable_name',\n",
       " 'ltable_brand',\n",
       " 'ltable_size',\n",
       " 'ltable_price',\n",
       " 'rtable_name',\n",
       " 'rtable_brand',\n",
       " 'rtable_size',\n",
       " 'rtable_price']"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "attrs_from_table = [attr for attr in labels.columns if attr not in [\"_id\", \"ltable_id\", \"rtable_id\", \"label\"]]\n",
    "attrs_from_table"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "0% [##############################] 100% | ETA: 00:00:00\n",
      "Total time elapsed: 00:00:03\n"
     ]
    }
   ],
   "source": [
    "train_feature_vecs = em.extract_feature_vecs(\n",
    "    train,\n",
    "    feature_table=feature_table,\n",
    "    attrs_before=attrs_from_table,\n",
    "    attrs_after=\"label\",\n",
    "    show_progress=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Train the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "attrs_to_be_excluded = []\n",
    "attrs_to_be_excluded.extend([\"_id\", \"ltable_id\", \"rtable_id\", \"label\"])\n",
    "attrs_to_be_excluded.extend(attrs_from_table)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "model.fit(\n",
    "    table=train_feature_vecs,\n",
    "    exclude_attrs=attrs_to_be_excluded,\n",
    "    target_attr=\"label\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Evaluate the model on the test set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate the feature vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "0% [##############################] 100% | ETA: 00:00:00\n",
      "Total time elapsed: 00:00:01\n"
     ]
    }
   ],
   "source": [
    "test_feature_vecs = em.extract_feature_vecs(\n",
    "    test,\n",
    "    feature_table=feature_table,\n",
    "    attrs_before=attrs_from_table,\n",
    "    attrs_after=\"label\",\n",
    "    show_progress=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Predict the label for the pairs in the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "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>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>predicted</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3583</th>\n",
       "      <td>altosight_352</td>\n",
       "      <td>altosight_805</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3367</th>\n",
       "      <td>altosight_526</td>\n",
       "      <td>altosight_541</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98</th>\n",
       "      <td>altosight_132</td>\n",
       "      <td>altosight_673</td>\n",
       "      <td>1</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5887</th>\n",
       "      <td>altosight_479</td>\n",
       "      <td>altosight_650</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4917</th>\n",
       "      <td>altosight_565</td>\n",
       "      <td>altosight_655</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>898</th>\n",
       "      <td>altosight_056</td>\n",
       "      <td>altosight_487</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1577</th>\n",
       "      <td>altosight_019</td>\n",
       "      <td>altosight_602</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7361</th>\n",
       "      <td>altosight_568</td>\n",
       "      <td>altosight_664</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1096</th>\n",
       "      <td>altosight_536</td>\n",
       "      <td>altosight_654</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5909</th>\n",
       "      <td>altosight_044</td>\n",
       "      <td>altosight_828</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2565 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "          ltable_id      rtable_id  predicted  weight\n",
       "3583  altosight_352  altosight_805          0     0.0\n",
       "3367  altosight_526  altosight_541          0     0.0\n",
       "98    altosight_132  altosight_673          1     1.0\n",
       "5887  altosight_479  altosight_650          0     0.0\n",
       "4917  altosight_565  altosight_655          0     0.0\n",
       "...             ...            ...        ...     ...\n",
       "898   altosight_056  altosight_487          0     0.0\n",
       "1577  altosight_019  altosight_602          0     0.0\n",
       "7361  altosight_568  altosight_664          0     0.0\n",
       "1096  altosight_536  altosight_654          0     0.0\n",
       "5909  altosight_044  altosight_828          0     0.0\n",
       "\n",
       "[2565 rows x 4 columns]"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_predictions = model.predict(\n",
    "    table=test_feature_vecs,\n",
    "    exclude_attrs=attrs_to_be_excluded,\n",
    "    target_attr=\"predicted\",\n",
    "    return_probs=True,\n",
    "    probs_attr=\"weight\",\n",
    "    append=True,\n",
    "    inplace=False\n",
    ")\n",
    "test_predictions[[\"ltable_id\", \"rtable_id\", \"predicted\", \"weight\"]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Compute precision, recall, and F-score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Precision: 43.182% | Recall: 61.29% | F-score: 50.667%'"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "test_pairs = {(pair[0], pair[1]) if pair[0] <= pair[1] else (pair[1], pair[0])\n",
    "              for pair in list(test_predictions[[\"ltable_id\", \"rtable_id\"]].itertuples(index=False, name=None))}\n",
    "predicted_matches = {(pair[0], pair[1]) if pair[0] <= pair[1] else (pair[1], pair[0])\n",
    "                     for pair in list(test_predictions[test_predictions[\"predicted\"] == 1][[\"ltable_id\", \"rtable_id\"]].itertuples(index=False, name=None))}\n",
    "matches = {(pair[0], pair[1]) if pair[0] <= pair[1] else (pair[1], pair[0])\n",
    "           for pair in list(gold[[\"l_id\", \"r_id\"]].itertuples(index=False, name=None))}.intersection(test_pairs)\n",
    "tp = len(predicted_matches.intersection(matches))\n",
    "fp = len(predicted_matches.difference(matches))\n",
    "fn = len(matches.difference(predicted_matches))\n",
    "precision = tp / (tp + fp)\n",
    "recall = tp / (tp + fn)\n",
    "f_score = (2 * precision * recall) / (precision + recall)\n",
    "\"Precision: \" + str(round(precision * 100, 3)) + \"% | Recall: \" + str(round(recall * 100, 3)) + \"% | F-score: \" + str(round(f_score * 100, 3)) + \"%\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load and prepare the candidate set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>altosight_114</td>\n",
       "      <td>altosight_702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>altosight_513</td>\n",
       "      <td>altosight_563</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>altosight_142</td>\n",
       "      <td>altosight_613</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>altosight_085</td>\n",
       "      <td>altosight_091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>altosight_297</td>\n",
       "      <td>altosight_718</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61559</th>\n",
       "      <td>61559</td>\n",
       "      <td>altosight_173</td>\n",
       "      <td>altosight_743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61560</th>\n",
       "      <td>61560</td>\n",
       "      <td>altosight_199</td>\n",
       "      <td>altosight_707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61561</th>\n",
       "      <td>61561</td>\n",
       "      <td>altosight_049</td>\n",
       "      <td>altosight_569</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61562</th>\n",
       "      <td>61562</td>\n",
       "      <td>altosight_204</td>\n",
       "      <td>altosight_684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61563</th>\n",
       "      <td>61563</td>\n",
       "      <td>altosight_089</td>\n",
       "      <td>altosight_657</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61564 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         _id      ltable_id      rtable_id\n",
       "0          0  altosight_114  altosight_702\n",
       "1          1  altosight_513  altosight_563\n",
       "2          2  altosight_142  altosight_613\n",
       "3          3  altosight_085  altosight_091\n",
       "4          4  altosight_297  altosight_718\n",
       "...      ...            ...            ...\n",
       "61559  61559  altosight_173  altosight_743\n",
       "61560  61560  altosight_199  altosight_707\n",
       "61561  61561  altosight_049  altosight_569\n",
       "61562  61562  altosight_204  altosight_684\n",
       "61563  61563  altosight_089  altosight_657\n",
       "\n",
       "[61564 rows x 3 columns]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "candidates = pd.read_csv(\"candidates.csv\")\n",
    "candidates = candidates.reset_index()\n",
    "candidates = candidates.rename(columns={\"index\": \"_id\", \"l_id\": \"ltable_id\", \"r_id\": \"rtable_id\"})\n",
    "candidates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "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>_id</th>\n",
       "      <th>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>ltable_name</th>\n",
       "      <th>ltable_brand</th>\n",
       "      <th>ltable_size</th>\n",
       "      <th>ltable_price</th>\n",
       "      <th>rtable_name</th>\n",
       "      <th>rtable_brand</th>\n",
       "      <th>rtable_size</th>\n",
       "      <th>rtable_price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>altosight_114</td>\n",
       "      <td>altosight_702</td>\n",
       "      <td>lexar jumpdrive p20 32gb - pendrive / memoria usb en fnac.es</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>lexar usb-stick 32 gb (ljdc20m-32gbbeu)</td>\n",
       "      <td>lexar</td>\n",
       "      <td>32.0</td>\n",
       "      <td>34.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>altosight_513</td>\n",
       "      <td>altosight_563</td>\n",
       "      <td>karta pami?ci sandisk microsdhc 16 gb extreme pro 95mb/s c10 uhs-i + adapter sd</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>16.0</td>\n",
       "      <td>103.00</td>\n",
       "      <td>sandisk cruzer extreme 16 gb usb-minne 3.0 - minneskort, usb-minne</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>16.0</td>\n",
       "      <td>189.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>altosight_142</td>\n",
       "      <td>altosight_613</td>\n",
       "      <td>sony sf8u, sd full size sd speicherkarte, 8 gb, 90 mb/s</td>\n",
       "      <td>sony</td>\n",
       "      <td>8.0</td>\n",
       "      <td>8.99</td>\n",
       "      <td>sony microsdxc 128 gb uhs-i u3 class 10 + sd adapter</td>\n",
       "      <td>sony</td>\n",
       "      <td>128.0</td>\n",
       "      <td>164.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>altosight_085</td>\n",
       "      <td>altosight_091</td>\n",
       "      <td>toshiba exceria pro - carte mã©moire flash - 128 go - sdxc uhs-ii</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>128.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>kingston - carte mã©moire flash - 16 go - microsdhc uhs-i</td>\n",
       "      <td>kingston</td>\n",
       "      <td>16.0</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>altosight_297</td>\n",
       "      <td>altosight_718</td>\n",
       "      <td>sandisk microsdhc 4gb class 4 speicherkarte (inkl. microsd zu sd adapter)</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>4.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>sony sdhc 16gb class 4</td>\n",
       "      <td>sony</td>\n",
       "      <td>16.0</td>\n",
       "      <td>590.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61559</th>\n",
       "      <td>61559</td>\n",
       "      <td>altosight_173</td>\n",
       "      <td>altosight_743</td>\n",
       "      <td>samsung galaxy a10 dual sim - 32gb, 2gb ram, 4g lte, blue</td>\n",
       "      <td>samsung</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1999.00</td>\n",
       "      <td>shop samsung galaxy a10 dual sim blue 32gb 4g lte online in egypt</td>\n",
       "      <td>samsung</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1999.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61560</th>\n",
       "      <td>61560</td>\n",
       "      <td>altosight_199</td>\n",
       "      <td>altosight_707</td>\n",
       "      <td>toshiba sdhc exceria pro 32go, silber</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>32.0</td>\n",
       "      <td>24.40</td>\n",
       "      <td>toshiba exceria pro m401, 16 gb</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>16.0</td>\n",
       "      <td>24.95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61561</th>\n",
       "      <td>61561</td>\n",
       "      <td>altosight_049</td>\n",
       "      <td>altosight_569</td>\n",
       "      <td>accessoires montres / bracelets connectã©s sandisk carte mã©moire sdxc sandisk extreme 128 go</td>\n",
       "      <td>sandisk</td>\n",
       "      <td>32.0</td>\n",
       "      <td>46.83</td>\n",
       "      <td>toshiba exceria n302 sdxc kort 128 gb - minneskort till kamera</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>128.0</td>\n",
       "      <td>699.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61562</th>\n",
       "      <td>61562</td>\n",
       "      <td>altosight_204</td>\n",
       "      <td>altosight_684</td>\n",
       "      <td>lexar jumpdrive c20c 128gb 128gb usb 3.0 (3.1 gen 1) conector usb tipo</td>\n",
       "      <td>lexar</td>\n",
       "      <td>128.0</td>\n",
       "      <td>63.99</td>\n",
       "      <td>kingston - flash-speicherkarte - 128 gb - uhs class 3 / class10 -</td>\n",
       "      <td>kingston</td>\n",
       "      <td>128.0</td>\n",
       "      <td>79.95</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61563</th>\n",
       "      <td>61563</td>\n",
       "      <td>altosight_089</td>\n",
       "      <td>altosight_657</td>\n",
       "      <td>transcend ultimate series - carte mã©moire flash - 64 go - sdxc uhs-ii</td>\n",
       "      <td>transcend</td>\n",
       "      <td>64.0</td>\n",
       "      <td>NaN</td>\n",
       "      <td>memoria toshiba exceria sdxc class 10 uhs-i class 3 64gb</td>\n",
       "      <td>toshiba</td>\n",
       "      <td>64.0</td>\n",
       "      <td>33.99</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61564 rows × 11 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "         _id      ltable_id      rtable_id  \\\n",
       "0          0  altosight_114  altosight_702   \n",
       "1          1  altosight_513  altosight_563   \n",
       "2          2  altosight_142  altosight_613   \n",
       "3          3  altosight_085  altosight_091   \n",
       "4          4  altosight_297  altosight_718   \n",
       "...      ...            ...            ...   \n",
       "61559  61559  altosight_173  altosight_743   \n",
       "61560  61560  altosight_199  altosight_707   \n",
       "61561  61561  altosight_049  altosight_569   \n",
       "61562  61562  altosight_204  altosight_684   \n",
       "61563  61563  altosight_089  altosight_657   \n",
       "\n",
       "                                                                                         ltable_name  \\\n",
       "0                                       lexar jumpdrive p20 32gb - pendrive / memoria usb en fnac.es   \n",
       "1                    karta pami?ci sandisk microsdhc 16 gb extreme pro 95mb/s c10 uhs-i + adapter sd   \n",
       "2                                            sony sf8u, sd full size sd speicherkarte, 8 gb, 90 mb/s   \n",
       "3                                  toshiba exceria pro - carte mã©moire flash - 128 go - sdxc uhs-ii   \n",
       "4                          sandisk microsdhc 4gb class 4 speicherkarte (inkl. microsd zu sd adapter)   \n",
       "...                                                                                              ...   \n",
       "61559                                      samsung galaxy a10 dual sim - 32gb, 2gb ram, 4g lte, blue   \n",
       "61560                                                          toshiba sdhc exceria pro 32go, silber   \n",
       "61561  accessoires montres / bracelets connectã©s sandisk carte mã©moire sdxc sandisk extreme 128 go   \n",
       "61562                         lexar jumpdrive c20c 128gb 128gb usb 3.0 (3.1 gen 1) conector usb tipo   \n",
       "61563                         transcend ultimate series - carte mã©moire flash - 64 go - sdxc uhs-ii   \n",
       "\n",
       "      ltable_brand  ltable_size  ltable_price  \\\n",
       "0            lexar         32.0           NaN   \n",
       "1          sandisk         16.0        103.00   \n",
       "2             sony          8.0          8.99   \n",
       "3          toshiba        128.0           NaN   \n",
       "4          sandisk          4.0           NaN   \n",
       "...            ...          ...           ...   \n",
       "61559      samsung          NaN       1999.00   \n",
       "61560      toshiba         32.0         24.40   \n",
       "61561      sandisk         32.0         46.83   \n",
       "61562        lexar        128.0         63.99   \n",
       "61563    transcend         64.0           NaN   \n",
       "\n",
       "                                                              rtable_name  \\\n",
       "0                                 lexar usb-stick 32 gb (ljdc20m-32gbbeu)   \n",
       "1      sandisk cruzer extreme 16 gb usb-minne 3.0 - minneskort, usb-minne   \n",
       "2                    sony microsdxc 128 gb uhs-i u3 class 10 + sd adapter   \n",
       "3               kingston - carte mã©moire flash - 16 go - microsdhc uhs-i   \n",
       "4                                                  sony sdhc 16gb class 4   \n",
       "...                                                                   ...   \n",
       "61559   shop samsung galaxy a10 dual sim blue 32gb 4g lte online in egypt   \n",
       "61560                                     toshiba exceria pro m401, 16 gb   \n",
       "61561      toshiba exceria n302 sdxc kort 128 gb - minneskort till kamera   \n",
       "61562   kingston - flash-speicherkarte - 128 gb - uhs class 3 / class10 -   \n",
       "61563            memoria toshiba exceria sdxc class 10 uhs-i class 3 64gb   \n",
       "\n",
       "      rtable_brand  rtable_size  rtable_price  \n",
       "0            lexar         32.0         34.99  \n",
       "1          sandisk         16.0        189.00  \n",
       "2             sony        128.0        164.50  \n",
       "3         kingston         16.0           NaN  \n",
       "4             sony         16.0        590.00  \n",
       "...            ...          ...           ...  \n",
       "61559      samsung          NaN       1999.00  \n",
       "61560      toshiba         16.0         24.95  \n",
       "61561      toshiba        128.0        699.00  \n",
       "61562     kingston        128.0         79.95  \n",
       "61563      toshiba         64.0         33.99  \n",
       "\n",
       "[61564 rows x 11 columns]"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "candidates = candidates.join(ds.set_index(\"_id\"), on=\"ltable_id\")\n",
    "candidates = candidates.rename(columns={attr: \"ltable_\" + attr for attr in attrs})\n",
    "candidates = candidates.join(ds.set_index(\"_id\"), on=\"rtable_id\")\n",
    "candidates = candidates.rename(columns={attr: \"rtable_\" + attr for attr in attrs})\n",
    "em.set_key(candidates, \"_id\")\n",
    "candidates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "candidates.to_csv(\"candidates_expanded.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Predict the labels for the candidates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Metadata file is not present in the given path; proceeding to read the csv file.\n"
     ]
    }
   ],
   "source": [
    "candidates = em.read_csv_metadata(\n",
    "    \"candidates_expanded.csv\",\n",
    "    key=\"_id\",\n",
    "    ltable=ds, rtable=ds,\n",
    "    fk_ltable=\"ltable_id\", fk_rtable=\"rtable_id\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generate the feature vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "0% [##############################] 100% | ETA: 00:00:00\n",
      "Total time elapsed: 00:00:27\n"
     ]
    }
   ],
   "source": [
    "pred_feature_vecs = em.extract_feature_vecs(\n",
    "    candidates,\n",
    "    feature_table=feature_table,\n",
    "    attrs_before=attrs_from_table,\n",
    "    show_progress=True\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Predict the labels for the candidates."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "attrs_to_be_excluded = []\n",
    "attrs_to_be_excluded.extend([\"_id\", \"ltable_id\", \"rtable_id\"])\n",
    "attrs_to_be_excluded.extend(attrs_from_table)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "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>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>predicted</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_114</td>\n",
       "      <td>altosight_702</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_513</td>\n",
       "      <td>altosight_563</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_142</td>\n",
       "      <td>altosight_613</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_085</td>\n",
       "      <td>altosight_091</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_297</td>\n",
       "      <td>altosight_718</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61559</th>\n",
       "      <td>altosight_173</td>\n",
       "      <td>altosight_743</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61560</th>\n",
       "      <td>altosight_199</td>\n",
       "      <td>altosight_707</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61561</th>\n",
       "      <td>altosight_049</td>\n",
       "      <td>altosight_569</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61562</th>\n",
       "      <td>altosight_204</td>\n",
       "      <td>altosight_684</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61563</th>\n",
       "      <td>altosight_089</td>\n",
       "      <td>altosight_657</td>\n",
       "      <td>0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61564 rows × 4 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           ltable_id      rtable_id  predicted  weight\n",
       "0      altosight_114  altosight_702          0     0.0\n",
       "1      altosight_513  altosight_563          0     0.0\n",
       "2      altosight_142  altosight_613          0     0.0\n",
       "3      altosight_085  altosight_091          0     0.0\n",
       "4      altosight_297  altosight_718          0     0.0\n",
       "...              ...            ...        ...     ...\n",
       "61559  altosight_173  altosight_743          0     0.0\n",
       "61560  altosight_199  altosight_707          0     0.0\n",
       "61561  altosight_049  altosight_569          0     0.0\n",
       "61562  altosight_204  altosight_684          0     0.0\n",
       "61563  altosight_089  altosight_657          0     0.0\n",
       "\n",
       "[61564 rows x 4 columns]"
      ]
     },
     "execution_count": 50,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions = model.predict(\n",
    "    table=pred_feature_vecs,\n",
    "    exclude_attrs=attrs_to_be_excluded,\n",
    "    target_attr=\"predicted\",\n",
    "    return_probs=True,\n",
    "    probs_attr=\"weight\",\n",
    "    append=True,\n",
    "    inplace=False\n",
    ")\n",
    "predictions[[\"ltable_id\", \"rtable_id\", \"predicted\", \"weight\"]]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Precision: 47.959% | Recall: 45.762% | F-score: 46.835%'"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predicted_matches = {(pair[0], pair[1]) if pair[0] <= pair[1] else (pair[1], pair[0])\n",
    "                     for pair in list(predictions[predictions[\"predicted\"] == 1][[\"ltable_id\", \"rtable_id\"]].itertuples(index=False, name=None))}\n",
    "matches = {(pair[0], pair[1]) if pair[0] <= pair[1] else (pair[1], pair[0])\n",
    "           for pair in list(gold[[\"l_id\", \"r_id\"]].itertuples(index=False, name=None))}\n",
    "tp = len(predicted_matches.intersection(matches))\n",
    "fp = len(predicted_matches.difference(matches))\n",
    "fn = len(matches.difference(predicted_matches))\n",
    "precision = tp / (tp + fp)\n",
    "recall = tp / (tp + fn)\n",
    "f_score = (2 * precision * recall) / (precision + recall)\n",
    "\"Precision: \" + str(round(precision * 100, 3)) + \"% | Recall: \" + str(round(recall * 100, 3)) + \"% | F-score: \" + str(round(f_score * 100, 3)) + \"%\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions[[\"ltable_id\", \"rtable_id\", \"weight\"]].to_csv(\"predictions.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Entity Clustering (PyJedAI)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It takes as input the similarity graph produced by Entity Matching and partitions it into a set of equivalence clusters, with every cluster corresponding to a distinct real-world object.\n",
    "\n",
    "https://github.com/AI-team-UoA/pyJedAI/blob/main/src/pyjedai/clustering.py"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "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>ltable_id</th>\n",
       "      <th>rtable_id</th>\n",
       "      <th>weight</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>altosight_114</td>\n",
       "      <td>altosight_702</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>altosight_513</td>\n",
       "      <td>altosight_563</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>altosight_142</td>\n",
       "      <td>altosight_613</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>altosight_085</td>\n",
       "      <td>altosight_091</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>altosight_297</td>\n",
       "      <td>altosight_718</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61559</th>\n",
       "      <td>altosight_173</td>\n",
       "      <td>altosight_743</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61560</th>\n",
       "      <td>altosight_199</td>\n",
       "      <td>altosight_707</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61561</th>\n",
       "      <td>altosight_049</td>\n",
       "      <td>altosight_569</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61562</th>\n",
       "      <td>altosight_204</td>\n",
       "      <td>altosight_684</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>61563</th>\n",
       "      <td>altosight_089</td>\n",
       "      <td>altosight_657</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61564 rows × 3 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           ltable_id      rtable_id  weight\n",
       "0      altosight_114  altosight_702     0.0\n",
       "1      altosight_513  altosight_563     0.0\n",
       "2      altosight_142  altosight_613     0.0\n",
       "3      altosight_085  altosight_091     0.0\n",
       "4      altosight_297  altosight_718     0.0\n",
       "...              ...            ...     ...\n",
       "61559  altosight_173  altosight_743     0.0\n",
       "61560  altosight_199  altosight_707     0.0\n",
       "61561  altosight_049  altosight_569     0.0\n",
       "61562  altosight_204  altosight_684     0.0\n",
       "61563  altosight_089  altosight_657     0.0\n",
       "\n",
       "[61564 rows x 3 columns]"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictions = pd.read_csv(\"predictions.csv\")\n",
    "predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "import networkx as nx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "id_mapping = {id_mapping[_id]: _id for _id in id_mapping.keys()}\n",
    "pairs_graph = nx.from_edgelist([(id_mapping[pair[0]], id_mapping[pair[1]], {\"weight\": pair[2]})\n",
    "                                for pair in predictions.itertuples(index=False, name=None)])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [],
   "source": [
    "from pyjedai.clustering import ConnectedComponentsClustering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "ec = ConnectedComponentsClustering()\n",
    "clusters = ec.process(pairs_graph, data, similarity_threshold=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "***************************************************************************************************************************\n",
      "                                         Μethod:  Connected Components Clustering\n",
      "***************************************************************************************************************************\n",
      "Method name: Connected Components Clustering\n",
      "Parameters: \n",
      "Runtime: 0.3840 seconds\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n",
      "Performance:\n",
      "\tPrecision:     19.76% \n",
      "\tRecall:        98.09%\n",
      "\tF1-score:      32.89%\n",
      "───────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────\n"
     ]
    }
   ],
   "source": [
    "_ = ec.evaluate(clusters)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": ".venv",
   "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": "3.9.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}