rp_x_metadata.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import re\n",
    "import scipy.stats\n",
    "import collections\n",
    "import seaborn\n",
    "import statsmodels.stats.proportion\n",
    "import matplotlib\n",
    "\n",
    "seaborn.set_style(\"darkgrid\")\n",
    "plt.rcParams.update({\"font.size\": 24})"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Read Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Filename                      \tAdvertiser                    \tDate                          \tVersion                       \tCountry                       \tDuration                      \n",
      "Description                   \tBrand/Product                 \tProductAdvertiser             \tTitle                         \tCreative ID                   \n",
      "Channel                       \tPublication                   \tWebsite                       \t!!MISSING_KEY!! CreativeLabel_Outlet_PaidSocial\tAd Format                     \n",
      "Endline                       \tvideo_key                     \tframes_analyzed               \t(female, [0, 15), [-inf, 1.1))\t(female, [0, 15), [1.1, 2.1)) \n",
      "(female, [0, 15), [2.1, 3.1)) \t(female, [0, 15), [3.1, 4.1)) \t(female, [0, 15), [4.1, 5.1)) \t(female, [0, 15), [5.1, 6.1)) \t(female, [0, 15), [6.1, 7.1)) \n",
      "(female, [0, 15), [7.1, 8.1)) \t(female, [0, 15), [8.1, 9.1)) \t(female, [15, 25), [-inf, 1.1))\t(female, [15, 25), [1.1, 2.1))\t(female, [15, 25), [2.1, 3.1))\n",
      "(female, [15, 25), [3.1, 4.1))\t(female, [15, 25), [4.1, 5.1))\t(female, [15, 25), [5.1, 6.1))\t(female, [15, 25), [6.1, 7.1))\t(female, [15, 25), [7.1, 8.1))\n",
      "(female, [15, 25), [8.1, 9.1))\t(female, [15, 25), [9.1, 10.1))\t(female, [25, 35), [-inf, 1.1))\t(female, [25, 35), [1.1, 2.1))\t(female, [25, 35), [2.1, 3.1))\n",
      "(female, [25, 35), [3.1, 4.1))\t(female, [25, 35), [4.1, 5.1))\t(female, [25, 35), [5.1, 6.1))\t(female, [25, 35), [6.1, 7.1))\t(female, [25, 35), [7.1, 8.1))\n",
      "(female, [25, 35), [8.1, 9.1))\t(female, [25, 35), [9.1, 10.1))\t(female, [35, 45), [-inf, 1.1))\t(female, [35, 45), [1.1, 2.1))\t(female, [35, 45), [2.1, 3.1))\n",
      "(female, [35, 45), [3.1, 4.1))\t(female, [35, 45), [4.1, 5.1))\t(female, [35, 45), [5.1, 6.1))\t(female, [35, 45), [6.1, 7.1))\t(female, [35, 45), [7.1, 8.1))\n",
      "(female, [35, 45), [8.1, 9.1))\t(female, [45, 55), [-inf, 1.1))\t(female, [45, 55), [1.1, 2.1))\t(female, [45, 55), [2.1, 3.1))\t(female, [45, 55), [3.1, 4.1))\n",
      "(female, [45, 55), [4.1, 5.1))\t(female, [45, 55), [5.1, 6.1))\t(female, [45, 55), [6.1, 7.1))\t(female, [45, 55), [7.1, 8.1))\t(female, [55, 65), [-inf, 1.1))\n",
      "(female, [55, 65), [1.1, 2.1))\t(female, [55, 65), [2.1, 3.1))\t(female, [55, 65), [3.1, 4.1))\t(female, [55, 65), [4.1, 5.1))\t(female, [55, 65), [6.1, 7.1))\n",
      "(female, [55, 65), [7.1, 8.1))\t(female, [65, 75), [1.1, 2.1))\t(male, [0, 15), [-inf, 1.1))  \t(male, [0, 15), [1.1, 2.1))   \t(male, [0, 15), [2.1, 3.1))   \n",
      "(male, [0, 15), [3.1, 4.1))   \t(male, [0, 15), [4.1, 5.1))   \t(male, [0, 15), [5.1, 6.1))   \t(male, [0, 15), [6.1, 7.1))   \t(male, [0, 15), [7.1, 8.1))   \n",
      "(male, [0, 15), [8.1, 9.1))   \t(male, [15, 25), [-inf, 1.1)) \t(male, [15, 25), [1.1, 2.1))  \t(male, [15, 25), [2.1, 3.1))  \t(male, [15, 25), [3.1, 4.1))  \n",
      "(male, [15, 25), [4.1, 5.1))  \t(male, [15, 25), [5.1, 6.1))  \t(male, [15, 25), [6.1, 7.1))  \t(male, [15, 25), [7.1, 8.1))  \t(male, [15, 25), [8.1, 9.1))  \n",
      "(male, [15, 25), [9.1, 10.1)) \t(male, [25, 35), [-inf, 1.1)) \t(male, [25, 35), [1.1, 2.1))  \t(male, [25, 35), [2.1, 3.1))  \t(male, [25, 35), [3.1, 4.1))  \n",
      "(male, [25, 35), [4.1, 5.1))  \t(male, [25, 35), [5.1, 6.1))  \t(male, [25, 35), [6.1, 7.1))  \t(male, [25, 35), [7.1, 8.1))  \t(male, [25, 35), [8.1, 9.1))  \n",
      "(male, [25, 35), [9.1, 10.1)) \t(male, [35, 45), [-inf, 1.1)) \t(male, [35, 45), [1.1, 2.1))  \t(male, [35, 45), [2.1, 3.1))  \t(male, [35, 45), [3.1, 4.1))  \n",
      "(male, [35, 45), [4.1, 5.1))  \t(male, [35, 45), [6.1, 7.1))  \t(male, [35, 45), [7.1, 8.1))  \t(male, [35, 45), [8.1, 9.1))  \t(male, [45, 55), [-inf, 1.1)) \n",
      "(male, [45, 55), [1.1, 2.1))  \t(male, [45, 55), [2.1, 3.1))  \t(male, [45, 55), [3.1, 4.1))  \t(male, [45, 55), [4.1, 5.1))  \t(male, [45, 55), [6.1, 7.1))  \n",
      "(male, [45, 55), [7.1, 8.1))  \t(male, [45, 55), [8.1, 9.1))  \t(male, [55, 65), [-inf, 1.1)) \t(male, [55, 65), [1.1, 2.1))  \t(male, [55, 65), [2.1, 3.1))  \n",
      "(male, [55, 65), [3.1, 4.1))  \t(male, [55, 65), [4.1, 5.1))  \t(male, [55, 65), [7.1, 8.1))  \t(male, [55, 65), [8.1, 9.1))  \t(male, [65, 75), [1.1, 2.1))  \n",
      "faces                         \tmst_scale_1                   \tmst_scale_2                   \tmst_scale_3                   \tmst_scale_4                   \n",
      "mst_scale_5                   \tmst_scale_6                   \tmst_scale_7                   \tmst_scale_8                   \tmst_scale_9                   \n",
      "mst_scale_10                  \tage_1                         \tage_2                         \tage_3                         \tage_4                         \n",
      "age_5                         \tage_6                         \tage_7                         \tmasculine_faces               \tfeminine_faces                \n",
      "video_key_to_match            \tbrand_category                \t\n",
      "total number of rows = 11577\n",
      "total number of videos = 10472\n",
      "total number of videos with RP output = 7648\n"
     ]
    }
   ],
   "source": [
    "# read data\n",
    "\n",
    "raw_df = pd.read_csv(\"raw_Mar31.csv\", index_col=None)\n",
    "for i, column in enumerate(raw_df.columns):\n",
    "    end = \"\\n\" if i % 5 == 0 and i > 0 else \"\\t\"\n",
    "    print(f\"{column:30s}\", end=end)\n",
    "print()\n",
    "print(f\"total number of rows = {raw_df.shape[0]}\")\n",
    "n_videos = raw_df[\"Filename\"].str.endswith(\".mp4\").sum()\n",
    "n_videos_rp = ((raw_df != -1).all(axis=1) & raw_df[\"Filename\"].str.endswith(\".mp4\")).sum()\n",
    "print(f\"total number of videos = {n_videos}\")\n",
    "print(f\"total number of videos with RP output = {n_videos_rp}\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Filter Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7648, 148)\n"
     ]
    }
   ],
   "source": [
    "# filter data\n",
    "\n",
    "data_df = raw_df.loc[(raw_df != -1).all(axis=1) & raw_df[\"Filename\"].str.endswith(\".mp4\")].copy()\n",
    "data_df.index = pd.RangeIndex(0, len(data_df))\n",
    "print(data_df.shape)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Check Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# check cross section bins count equals aggregates\n",
    "\n",
    "def describe(row):\n",
    "    print(row[\"Creative ID\"])\n",
    "    rp_columns = [column for column in row.index if re.match(r\"\\((fe)?male\", column)]\n",
    "    for column in rp_columns:\n",
    "        if row[column] != 0:\n",
    "            print(f\"\\t{column} : {row[column]}\")\n",
    "    print(f\"faces : {row.faces}\")\n",
    "    print(f\"gender : male = {row.masculine_faces}, female = {row.feminine_faces}\")\n",
    "    print(\"age : \")\n",
    "    for j in range(7):\n",
    "        count = row[f\"age_{j + 1}\"]\n",
    "        end = \", \" if j < 6 else \"\\n\"\n",
    "        print(f\"age_{j + 1} = {count}\", end=end)\n",
    "    print(\"monk skin tone : \")\n",
    "    for k in range(10):\n",
    "        count = row[f\"mst_scale_{k + 1}\"]\n",
    "        end = \", \" if k < 9 else \"\\n\"\n",
    "        print(f\"mst_scale_{k + 1} = {count}\", end=end)\n",
    "\n",
    "for _, row in data_df.iterrows():\n",
    "    rp_data = np.zeros((2, 7, 10), dtype=int)\n",
    "    for i, gender in enumerate([\"male\", \"female\"]):\n",
    "        for j in range(7):\n",
    "            for k in range(10):\n",
    "                age_lb = 0 if j == 0 else 10 * j + 5\n",
    "                age_ub = 10 * j + 15\n",
    "                mst_lb = \"-inf\" if k == 0 else k + 0.1\n",
    "                mst_ub = k + 1.1\n",
    "                column_name = f\"({gender}, [{age_lb}, {age_ub}), [{mst_lb}, {mst_ub}))\"\n",
    "                if column_name in row:\n",
    "                    rp_data[i, j, k] = row[column_name]\n",
    "    assert rp_data.sum() == row[\"faces\"]\n",
    "    assert rp_data[0].sum() == row[\"masculine_faces\"]\n",
    "    assert rp_data[1].sum() == row[\"feminine_faces\"]\n",
    "    for j in range(7):\n",
    "        assert rp_data[:,j].sum() == row[f\"age_{j + 1}\"]\n",
    "    # for k in range(10):\n",
    "    #     assert rp_data[:, :, k].sum() == row[f\"mst_scale_{k + 1}\"]"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Metadata Distributions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/var/folders/r0/x3cynkz966z2t0k946xpqqsc0000gn/T/ipykernel_1789/3934675033.py:24: FutureWarning: The default value of regex will change from True to False in a future version.\n",
      "  data_df[\"brand\"] = data_df[\"brand\"].str.replace(\"^MAYBELLINE$\", \"MNY\")\n"
     ]
    }
   ],
   "source": [
    "def target_gender(desc):\n",
    "    desc = str(desc)\n",
    "    f = re.search(r\"(^|\\s)(females?)|(wom(a|e)n)(\\s|$)\", desc.lower()) is not None\n",
    "    m = re.search(r\"(^|\\s)((?<!fe)males?)|((?<!wo)m(a|e)n)(\\s|$)\", desc.lower()) is not None\n",
    "    u = re.search(r\"(^|\\s)unisex(\\s|$)\", desc.lower()) is not None\n",
    "    if f and not m and not u:\n",
    "        return \"female\"\n",
    "    elif m and not f:\n",
    "        return \"male\"\n",
    "    elif u and not f and not m:\n",
    "        return \"unisex\"\n",
    "    else:\n",
    "        return \"unspecified\"\n",
    "\n",
    "# metadata\n",
    "\n",
    "data_df[\"country\"] = data_df[\"Country\"]\n",
    "data_df[\"brand\"] = data_df[\"ProductAdvertiser\"]\n",
    "data_df[\"product type\"] = data_df[\"brand_category\"]\n",
    "# data_df.loc[data_df[\"Brand/Product\"].apply(lambda x: re.search(\"(\\s|^)((toiletries)|(toilet))(\\s|$)\", x) is not None), \n",
    "#             \"product type\"] = \"toiletries\"\n",
    "\n",
    "# brand: MNY = MAYBELLINE\n",
    "data_df[\"brand\"] = data_df[\"brand\"].str.replace(\"^MAYBELLINE$\", \"MNY\")\n",
    "data_df[\"product target gender\"] = data_df[\"Brand/Product\"].apply(target_gender)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/tmp/ipykernel_34980/3578326182.py:10: FutureWarning: The default value of regex will change from True to False in a future version.\n",
      "  data_df[\"brand\"] = data_df[\"brand\"].str.replace(\"^MAYBELLINE$\", \"MNY\")\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "country\n",
      "Germany    : 3202 (41.9%)\tUK         : 1715 (22.4%)\tFrance     :  817 (10.7%)\tItaly      :  806 (10.5%)\tUSA        :  701 ( 9.2%)\tSpain      :  227 ( 3.0%)\n",
      "Brazil     :   54 ( 0.7%)\tIndia      :   49 ( 0.6%)\tMexico     :   46 ( 0.6%)\tIndonesia  :   30 ( 0.4%)\tCanada     :    1 ( 0.0%)\n",
      "\n",
      "\n",
      "brand\n",
      "GARNIER    : 1825 (23.9%)\tOAP        :  880 (11.5%)\tMNY        :  754 ( 9.9%)\tLRP        :  631 ( 8.3%)\tYSL        :  613 ( 8.0%)\tVICHY      :  511 ( 6.7%)\n",
      "LANCOME    :  504 ( 6.6%)\tKIEHLS     :  429 ( 5.6%)\tARMANI     :  337 ( 4.4%)\tKERASTASE  :  335 ( 4.4%)\tNYX        :  286 ( 3.7%)\n",
      "CERAVE     :  239 ( 3.1%)\tLOP        :  110 ( 1.4%)\tPRADA      :   84 ( 1.1%)\tBIOTHERM   :   80 ( 1.0%)\tREDKEN     :   30 ( 0.4%)\n",
      "\n",
      "\n",
      "product type\n",
      "makeup     : 2897 (37.9%)\tskincare   : 2429 (31.8%)\thaircare   : 1640 (21.4%)\tfragrance  :  682 ( 8.9%)\t\n",
      "\n",
      "product target gender\n",
      "unspecified : 4844 (63.3%)\tfemale     : 2155 (28.2%)\tmale       :  460 ( 6.0%)\tunisex     :  189 ( 2.5%)\t"
     ]
    }
   ],
   "source": [
    "def plural(metadata):\n",
    "    if metadata == \"country\":\n",
    "        return \"countries\"\n",
    "    elif metadata == \"brand\":\n",
    "        return \"brands\"\n",
    "    elif metadata == \"product type\":\n",
    "        return \"product types\"\n",
    "    elif metadata == \"product target gender\":\n",
    "        return \"product target gender categories\"\n",
    "    else:\n",
    "        return metadata + \"s\"\n",
    "\n",
    "def plot_metadata_distribution(dist, metadata):\n",
    "    plt.figure(figsize=(22, 10))\n",
    "    x = [item[0] for item in dist]\n",
    "    y = [item[1] for item in dist]\n",
    "    total = sum(y)\n",
    "    yper = [100*yy/total for yy in y]\n",
    "    xx = np.arange(len(x))\n",
    "    bc = plt.bar(xx, y, width=0.8, edgecolor=\"black\", linewidth=2)\n",
    "    plt.bar_label(bc, [f\"{yy:.1f}%\" for yy in yper], padding=10, rotation=0)\n",
    "    bottom, top = plt.ylim()\n",
    "    plt.ylim(bottom, top + 50)\n",
    "    xrotation = 90 if len(x) > 12 else 0\n",
    "    plt.xticks(xx, x, rotation=xrotation, ha=\"center\", fontsize=\"small\")\n",
    "    plt.yticks(fontsize=\"small\")\n",
    "    plt.xlabel(metadata.upper())\n",
    "    plt.ylabel(\"NUMBER OF ADS\")\n",
    "    metadata_in_title = plural(metadata)\n",
    "    plt.title(f\"DISTRIBUTION OF ADS ACROSS {metadata_in_title.upper()}\")\n",
    "    plt.savefig(f\"plots/metadata/{metadata}.png\", bbox_inches=\"tight\", pad_inches=0, dpi=300)\n",
    "    plt.close()\n",
    "\n",
    "country_dist = sorted(collections.Counter(data_df[\"country\"]).items(), key=lambda x: x[1], reverse=True)\n",
    "brand_dist = sorted(collections.Counter(data_df[\"brand\"]).items(), key=lambda x: x[1], reverse=True)\n",
    "producttype_dist = sorted(collections.Counter(data_df[\"product type\"]).items(), key=lambda x: x[1], reverse=True)\n",
    "targetgender_dist = sorted(collections.Counter(data_df[\"product target gender\"]).items(), key=lambda x: x[1], \n",
    "                           reverse=True)\n",
    "total = len(data_df)\n",
    "\n",
    "print(\"country\")\n",
    "for i, (country, count) in enumerate(country_dist):\n",
    "    per = 100*count/total\n",
    "    end = \"\\n\" if i % 5 == 0 and i > 0 else \"\\t\"\n",
    "    print(f\"{country:10s} : {count:4d} ({per:4.1f}%)\", end=end)\n",
    "plot_metadata_distribution(country_dist, \"country\")\n",
    "print(\"\\n\")\n",
    "\n",
    "print(\"brand\")\n",
    "for i, (brand, count) in enumerate(brand_dist):\n",
    "    per = 100*count/total\n",
    "    end = \"\\n\" if i % 5 == 0 and i > 0 else \"\\t\"\n",
    "    print(f\"{brand:10s} : {count:4d} ({per:4.1f}%)\", end=end)\n",
    "plot_metadata_distribution(brand_dist, \"brand\")\n",
    "print(\"\\n\")\n",
    "\n",
    "print(\"product type\")\n",
    "for i, (producttype, count) in enumerate(producttype_dist):\n",
    "    per = 100*count/total\n",
    "    end = \"\\n\" if i % 5 == 0 and i > 0 else \"\\t\"\n",
    "    print(f\"{producttype:10s} : {count:4d} ({per:4.1f}%)\", end=end)\n",
    "plot_metadata_distribution(producttype_dist, \"product type\")\n",
    "print(\"\\n\")\n",
    "\n",
    "print(\"product target gender\")\n",
    "for i, (targetgender, count) in enumerate(targetgender_dist):\n",
    "    per = 100*count/total\n",
    "    end = \"\\n\" if i % 5 == 0 and i > 0 else \"\\t\"\n",
    "    print(f\"{targetgender:10s} : {count:4d} ({per:4.1f}%)\", end=end)\n",
    "plot_metadata_distribution(targetgender_dist, \"product target gender\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "def metadata_x_metadata_stacked(data_df, primary_metadata, secondary_metadata, figsize=(22, 10)):\n",
    "    assert primary_metadata in [\"brand\", \"country\", \"product type\", \"product target gender\"]\n",
    "    assert secondary_metadata in [\"brand\", \"country\", \"product type\", \"product target gender\"]\n",
    "    n_primary_metadata_cats = data_df[primary_metadata].unique().size\n",
    "    primary_metadata_dist = collections.Counter(data_df[primary_metadata].tolist())\n",
    "    primary_metadata_cats = sorted(primary_metadata_dist.keys(), key=lambda cat: primary_metadata_dist[cat], \n",
    "                                   reverse=True)\n",
    "    n_secondary_metadata_cats = data_df[secondary_metadata].unique().size\n",
    "    secondary_metadata_dist = collections.Counter(data_df[secondary_metadata].tolist())\n",
    "    secondary_metadata_cats = sorted(secondary_metadata_dist.keys(), key=lambda cat: secondary_metadata_dist[cat], \n",
    "                                   reverse=True)\n",
    "    metadata_x_metadata_arr = np.zeros((n_primary_metadata_cats, n_secondary_metadata_cats), dtype=int)\n",
    "    for (primary_metadata_cat, secondary_metadata_cat), df in data_df.groupby([primary_metadata, secondary_metadata]):\n",
    "        i = primary_metadata_cats.index(primary_metadata_cat)\n",
    "        j = secondary_metadata_cats.index(secondary_metadata_cat)\n",
    "        metadata_x_metadata_arr[i, j] = len(df)\n",
    "    metadata_x_metadata_arr = 100 * metadata_x_metadata_arr / metadata_x_metadata_arr.sum(axis=1, keepdims=True)\n",
    "    bar_width = 0.8\n",
    "    bottom = np.zeros(n_primary_metadata_cats, dtype=float)\n",
    "    x = np.arange(n_primary_metadata_cats)\n",
    "    plt.figure(figsize=figsize)\n",
    "    for j, secondary_metadata_cat in enumerate(secondary_metadata_cats):\n",
    "        height = metadata_x_metadata_arr[:, j]\n",
    "        plt.bar(x, height, width=bar_width, bottom=bottom, label=secondary_metadata_cat, edgecolor=\"black\", lw=2)\n",
    "        bottom += height\n",
    "    plt.legend(bbox_to_anchor=(1, 1))\n",
    "    xrotation = 90 if n_primary_metadata_cats > 12 else 0\n",
    "    plt.xticks(x, primary_metadata_cats, fontsize=\"small\", rotation=xrotation)\n",
    "    plt.xlabel(primary_metadata)\n",
    "    plt.ylabel(\"PERCENTAGE OF ADS\")\n",
    "    plt.title(f\"{secondary_metadata.upper()} DISTRIBUTION OF {primary_metadata.upper()} ADS\")\n",
    "    plt.savefig(f\"plots/metadata_x_metadata/{primary_metadata}_x_{secondary_metadata}.png\", bbox_inches=\"tight\", \n",
    "                pad_inches=0, dpi=300)\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "metadata_x_metadata_stacked(data_df, \"brand\", \"product type\")\n",
    "metadata_x_metadata_stacked(data_df, \"country\", \"product type\")\n",
    "metadata_x_metadata_stacked(data_df, \"brand\", \"product target gender\")\n",
    "metadata_x_metadata_stacked(data_df, \"country\", \"product target gender\")\n",
    "metadata_x_metadata_stacked(data_df, \"product type\", \"product target gender\")\n",
    "metadata_x_metadata_stacked(data_df, \"brand\", \"country\", figsize=(22, 15))\n",
    "metadata_x_metadata_stacked(data_df, \"country\", \"brand\", figsize=(22, 15))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# rp_data rows x gender x age x skin tone\n",
    "\n",
    "gender_cats = [\"male\", \"female\"]\n",
    "age_cats = [(0, 15), (15, 25), (25, 35), (35, 45), (45, 55), (55, 65), (65, 75)]\n",
    "age_cats2 = [\"<15y\", \"15y-25y\", \"25y-35y\", \"35y-45y\", \"45-75y\"]\n",
    "skintone_cats = [(-np.inf, 1.1), (1.1, 2.1), (2.1, 3.1), (3.1, 4.1), (4.1, 5.1),\n",
    "                 (5.1, 6.1), (6.1, 7.1), (7.1, 8.1), (8.1, 9.1), (9.1, 10.1)]\n",
    "skintone_cats2 = [\"light\", \"medium\", \"dark\"]\n",
    "\n",
    "rp_data = np.zeros((len(data_df), len(gender_cats), len(age_cats), len(skintone_cats)), dtype=int)\n",
    "\n",
    "for i, row in data_df.iterrows():\n",
    "    for j, gender in enumerate(gender_cats):\n",
    "        for k, age_bin in enumerate(age_cats):\n",
    "            age_lb, age_ub = age_bin\n",
    "            for l, skintone_bin in enumerate(skintone_cats):\n",
    "                skintone_lb, skintone_ub = skintone_bin\n",
    "                column = f\"({gender}, [{age_lb}, {age_ub}), [{skintone_lb}, {skintone_ub}))\"\n",
    "                if column in row.index:\n",
    "                    rp_data[i, j, k, l] += row[column]\n",
    "\n",
    "rp_data2 = np.zeros((len(data_df), len(gender_cats), len(age_cats2), len(skintone_cats2)), dtype=int)\n",
    "rp_data2[:, :, :4, 0] = rp_data[:, :, :4, :3].sum(axis=3)\n",
    "rp_data2[:, :, :4, 1] = rp_data[:, :, :4, 3:7].sum(axis=3)\n",
    "rp_data2[:, :, :4, 2] = rp_data[:, :, :4, 7:].sum(axis=3)\n",
    "rp_data2[:, :, 4, 0] = rp_data[:, :, 4:, :3].sum(axis=(2,3))\n",
    "rp_data2[:, :, 4, 1] = rp_data[:, :, 4:, 3:7].sum(axis=(2, 3))\n",
    "rp_data2[:, :, 4, 2] = rp_data[:, :, 4:, 7:].sum(axis=(2,3))"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Skintone 10"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7648, 10)\n"
     ]
    }
   ],
   "source": [
    "skintone_data = rp_data.sum(axis=(1,2))\n",
    "print(skintone_data.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = skintone_data/skintone_data.sum(axis=1, keepdims=True)\n",
    "y = np.mean(100*y, axis=0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 2200x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "x = np.arange(1, 11)\n",
    "plt.figure(figsize=(22, 10))\n",
    "bc = plt.bar(x, y, width=0.8, color=[\"b\", \"b\", \"b\", \"g\", \"g\", \"g\", \"g\", \"r\", \"r\", \"r\"])\n",
    "plt.bar_label(bc, [f\"{yy:.1f}%\" for yy in y], padding=10)\n",
    "plt.title(\"Screentime distribution of skintone bins\")\n",
    "plt.xlabel(\"MST skintone bins\")\n",
    "plt.ylabel(\"Average screentime percentage\")\n",
    "plt.xticks(x, x)\n",
    "plt.ylim(0, 100)\n",
    "plt.legend([bc.patches[0], bc.patches[3], bc.patches[7]], [\"light\", \"medium\", \"dark\"])\n",
    "plt.savefig(\"plots/skintone10.png\", bbox_inches=\"tight\", pad_inches=0, dpi=300)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7648, 3)\n"
     ]
    }
   ],
   "source": [
    "skintone_data2 = rp_data2.sum(axis=(1,2))\n",
    "print(skintone_data2.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([81.39517347, 12.01032912,  6.59449741])"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y2 = skintone_data2/skintone_data2.sum(axis=1, keepdims=True)\n",
    "y2 = np.mean(100*y2, axis=0)\n",
    "y2"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Skintone 10 vs Brand, Country"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "((7648, 2, 7, 10), (7648, 2), 16, 11)"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "rp_data.shape, data_df[[\"brand\", \"country\"]].shape, data_df[\"brand\"].unique().size, data_df[\"country\"].unique().size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "skintone_data = rp_data.sum(axis=(1,2))\n",
    "skintone_percentage = 100*(skintone_data/skintone_data.sum(axis=1, keepdims=True))\n",
    "\n",
    "brand_arr = []\n",
    "for brand in sorted(data_df[\"brand\"].unique()):\n",
    "    row = [brand, (data_df[\"brand\"] == brand).sum()]\n",
    "    for i in range(10):\n",
    "        avg = skintone_percentage[data_df[\"brand\"] == brand, i].mean()\n",
    "        row.append(avg)\n",
    "    brand_arr.append(row)\n",
    "brand_skintone_df = pd.DataFrame(brand_arr, columns=[\"brand\", \"count\"] + [f\"MST {i}\" for i in range(1, 11)])\n",
    "brand_skintone_df.to_csv(\"skintone_vs_brand.csv\", index=False)\n",
    "\n",
    "country_arr = []\n",
    "for country in sorted(data_df[\"country\"].unique()):\n",
    "    row = [country, (data_df[\"country\"] == country).sum()]\n",
    "    for i in range(10):\n",
    "        avg = skintone_percentage[data_df[\"country\"] == country, i].mean()\n",
    "        row.append(avg)\n",
    "    country_arr.append(row)\n",
    "country_skintone_df = pd.DataFrame(country_arr, columns=[\"country\", \"count\"] + [f\"MST {i}\" for i in range(1, 11)])\n",
    "country_skintone_df.to_csv(\"skintone_vs_country.csv\", index=False)\n",
    "\n",
    "brand_x_country_arr = []\n",
    "for brand in sorted(data_df[\"brand\"].unique()):\n",
    "    for country in sorted(data_df[\"country\"].unique()):\n",
    "        index = (data_df[\"brand\"] == brand) & (data_df[\"country\"] == country)\n",
    "        count = index.sum()\n",
    "        if count:\n",
    "            row = [brand, country, count]\n",
    "            for i in range(10):\n",
    "                avg = skintone_percentage[index, i].mean()\n",
    "                row.append(avg)\n",
    "            brand_x_country_arr.append(row)\n",
    "brand_x_country_df = pd.DataFrame(brand_x_country_arr, columns=[\"brand\", \"country\", \"count\"]\n",
    "                                   + [f\"MST {i}\" for i in range(1, 11)])\n",
    "brand_x_country_df.to_csv(\"skintone_vs_brand_x_country.csv\", index=False)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Anti-Aging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "26\n",
      "38\n",
      "7\n",
      "92\n",
      "990\n",
      "880\n",
      "504\n",
      "26\n",
      "38\n",
      "7\n",
      "59\n"
     ]
    }
   ],
   "source": [
    "f1 = data_df[\"Brand/Product\"].str.lower().str.contains(r\"age\\Wperfect\")\n",
    "f2 = data_df[\"Brand/Product\"].str.lower().str.contains(r\"l\\\\?'?oreal\")\n",
    "f3 = (data_df[\"brand\"] == \"OAP\")\n",
    "f4 = data_df[\"Brand/Product\"].str.lower().str.contains(r\"revitalift\\Wlaser\")\n",
    "f5 = data_df[\"Brand/Product\"].str.lower().str.contains(r\"revitalift\\Wfiller\")\n",
    "f6 = data_df[\"Brand/Product\"].str.lower().str.contains(r\"gene|ifique\")\n",
    "f7 = (data_df[\"brand\"] == \"LANCOME\")\n",
    "print(f1.sum())\n",
    "print(f4.sum())\n",
    "print(f5.sum())\n",
    "print(f6.sum())\n",
    "print(f2.sum())\n",
    "print(f3.sum())\n",
    "print(f7.sum())\n",
    "print((f1 & f2 & f3).sum())\n",
    "print((f4 & f2 & f3).sum())\n",
    "print((f5 & f2 & f3).sum())\n",
    "print((f6 & f7).sum())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Counter({0: 5217, 2: 2301, 1: 130})"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "f8 = ((f1 | f4 | f5) & f2 & f3) | (f6 & f7)\n",
    "f9 = ~f8 & (data_df[\"product type\"] == \"skincare\")\n",
    "data_df[\"anti-aging\"] = 0\n",
    "data_df.loc[f8, \"anti-aging\"] = 1\n",
    "data_df.loc[f9 , \"anti-aging\"] = 2\n",
    "collections.Counter(data_df[\"anti-aging\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7648, 5)\n",
      "(7648, 2)\n",
      "True\n"
     ]
    }
   ],
   "source": [
    "rp_data3 = rp_data2.sum(axis=(1,3))\n",
    "print(rp_data3.shape)\n",
    "rp_data3[:,0] = rp_data3[:,:3].sum(axis=1)\n",
    "rp_data3[:,3] = rp_data3[:,3:].sum(axis=1)\n",
    "rp_data3 = rp_data3[:,[0,3]]\n",
    "print(rp_data3.shape)\n",
    "print((rp_data2.sum(axis=(1,2,3)) == rp_data3.sum(axis=1)).all())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(7648, 2)\n"
     ]
    }
   ],
   "source": [
    "antiage_percentage = 100 * rp_data3/rp_data3.sum(axis=1, keepdims=True)\n",
    "print(antiage_percentage.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[92.91565164,  7.08434836],\n",
       "       [88.36937044, 11.63062956]])"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "np.vstack((antiage_percentage[data_df[\"anti-aging\"] == 2].mean(axis=0), \n",
    "           antiage_percentage[data_df[\"anti-aging\"] == 1].mean(axis=0)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(130,) 88.36937044112939 (2301,) 92.91565164040503\n",
      "0.018718255480776512\n"
     ]
    }
   ],
   "source": [
    "x = antiage_percentage[data_df[\"anti-aging\"] == 1, 0] # screen time of young people (<35) in ads of anti-aging products\n",
    "y = antiage_percentage[data_df[\"anti-aging\"] == 2, 0] # screen time of young people (<35) in ads of not anti-aging products\n",
    "print(x.shape, x.mean(), y.shape, y.mean())\n",
    "res = scipy.stats.ttest_ind(x, y, equal_var=False, alternative=\"less\")\n",
    "print(res.pvalue)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Demographic Identity vs Metadata"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# one metadata vs one rp\n",
    "\n",
    "def metadata_x_rp(df, rp_data, metadata, rp, confidence=0.05, condition=None):\n",
    "    assert rp in [\"gender\", \"age\", \"skin tone\"]\n",
    "    assert metadata in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    \n",
    "    if condition is None:\n",
    "        plot_filename = f\"plots/rp_x_metadata/{rp}_x_{metadata}.png\"\n",
    "        text_filename = f\"text/rp_x_metadata/{rp}_x_{metadata}.txt\"\n",
    "    else:\n",
    "        plot_filename = f\"plots/rp_x_metadata_x_metadata/{rp}_x_{metadata}_{condition}.png\"\n",
    "        text_filename = f\"text/rp_x_metadata_x_metadata/{rp}_x_{metadata}_{condition}.txt\"\n",
    "    \n",
    "    metadata_cat_to_count = collections.Counter(df[metadata])\n",
    "    metadata_cats = [metadata_cat for metadata_cat, count in metadata_cat_to_count.items() if count >= 30]\n",
    "    removed_metadata_cats = [x for x in df[metadata].unique() if x not in metadata_cats]\n",
    "    if len(metadata_cats) <= 1:\n",
    "        return\n",
    "    \n",
    "    title = f\"{rp} vs {metadata}\"\n",
    "    if condition is not None:\n",
    "        title += f\" {condition}\"\n",
    "    \n",
    "    \n",
    "    print(title)\n",
    "    print(f\"{metadata} categories : {metadata_cats}\")\n",
    "    print(f\"removed categories : {removed_metadata_cats}\")\n",
    "    \n",
    "    fw = open(text_filename, \"w\")\n",
    "    fw.write(f\"{title}\\n\")\n",
    "    fw.write(f\"{metadata} cats = {metadata_cats}\\n\")\n",
    "    fw.write(f\"removed {metadata} cats = {removed_metadata_cats}\\n\")\n",
    "    \n",
    "    if rp == \"gender\":\n",
    "        rp_cats = gender_cats\n",
    "    elif rp == \"age\":\n",
    "        rp_cats = age_cats2\n",
    "    else:\n",
    "        rp_cats = skintone_cats2\n",
    "    \n",
    "    rp_arr = np.zeros((df[metadata].isin(metadata_cats).sum(), len(rp_cats)))\n",
    "    metadata_arr = []\n",
    "    j = 0\n",
    "    \n",
    "    for i, row in df.iterrows():\n",
    "        metadata_value = row[metadata]\n",
    "        if metadata_value in metadata_cats:\n",
    "            metadata_arr.append(metadata_value)\n",
    "            if rp == \"gender\":\n",
    "                rp_arr[j] = 100 * rp_data[i].sum(axis=(1, 2))/rp_data[i].sum()\n",
    "            elif rp == \"age\":\n",
    "                rp_arr[j] = 100 * rp_data[i].sum(axis=(0, 2))/rp_data[i].sum()\n",
    "            else:\n",
    "                rp_arr[j] = 100 * rp_data[i].sum(axis=(0, 1))/rp_data[i].sum()\n",
    "            j += 1\n",
    "    \n",
    "    metadata_arr = np.array(metadata_arr)\n",
    "    most_common_metadata_cat = rp_arr.sum(axis=0).argmax()\n",
    "    metadata_cats = sorted(metadata_cats, key=lambda x: rp_arr[:, most_common_metadata_cat][metadata_arr == x].mean(), \n",
    "                           reverse=True)\n",
    "    \n",
    "    plt.figure(figsize=(22, 10))\n",
    "    x = np.arange(len(metadata_cats))\n",
    "    group_width = 0.8\n",
    "    bar_width = group_width/len(rp_cats)\n",
    "\n",
    "    for i, rp_cat in enumerate(rp_cats):\n",
    "        rp_cat_arr = rp_arr[:, i]\n",
    "        y = [rp_cat_arr[metadata_arr == metadata_cat].mean() for metadata_cat in metadata_cats]\n",
    "        bc = plt.bar(x + i * bar_width, y, width=bar_width, label=str(rp_cat), edgecolor=\"black\", linewidth=2)\n",
    "        plt.bar_label(bc, [f\"{yy:.1f}%\" for yy in y], padding=5, rotation=90, fontsize=\"small\")\n",
    "    \n",
    "    bottom, top = plt.ylim()\n",
    "    plt.ylim(bottom, top + 10)\n",
    "    plt.ylabel(\"AVERAGE SCREEN TIME\")\n",
    "    xrotation = 90 if len(x) > 10 else 0\n",
    "    plt.xticks(x + bar_width*(len(rp_cats) - 1)/2, metadata_cats, ha=\"center\", rotation=xrotation)\n",
    "    plt.yticks(fontsize=\"small\")\n",
    "    plt.xlabel(metadata.upper())\n",
    "    condition_title = \" (\" + condition + \")\" if condition is not None else \"\"\n",
    "    plt.title(f\"DISTRIBUTION OF {rp.upper()} SCREEN TIME ACROSS {plural(metadata).upper()}{condition_title}\")\n",
    "    plt.legend()\n",
    "    plt.savefig(plot_filename, bbox_inches=\"tight\", pad_inches=0, dpi=300)\n",
    "    plt.close()\n",
    "\n",
    "    for i, rp_cat in enumerate(rp_cats):\n",
    "        rp_cat_arr = rp_arr[:, i]\n",
    "        metadata_to_rp_arr = {metadata_cat:rp_cat_arr[metadata_arr == metadata_cat] for metadata_cat in metadata_cats}\n",
    "        metadata_sorted_by_mean = sorted(metadata_cats, key=lambda x: metadata_to_rp_arr[x].mean(), reverse=True)\n",
    "        pvalues = []\n",
    "        \n",
    "        for j in range(len(metadata_cats) - 1):\n",
    "            cat1, cat2 = metadata_sorted_by_mean[j], metadata_sorted_by_mean[j + 1]\n",
    "            arr1, arr2 = metadata_to_rp_arr[cat1], metadata_to_rp_arr[cat2]\n",
    "            _, pvalue = scipy.stats.ttest_ind(arr1, arr2, equal_var=False, alternative=\"greater\")\n",
    "            pvalues.append(pvalue)\n",
    "        significant = np.array(pvalues)/(len(metadata_cats) - 1) < confidence\n",
    "        print(f\"{rp_cat} : \", end=\"\")\n",
    "        fw.write(f\"{rp_cat}\\n\")\n",
    "        \n",
    "        for j in range(len(metadata_cats) - 1):\n",
    "            cat = metadata_sorted_by_mean[j]\n",
    "            sign = \">\" if significant[j] else \"=\"\n",
    "            print(f\"{cat} {sign} \", end=\"\")\n",
    "            fw.write(f\"{cat} {sign} \")\n",
    "        print(metadata_sorted_by_mean[-1])\n",
    "        fw.write(f\"{metadata_sorted_by_mean[-1]}\\n\")\n",
    "    print()\n",
    "    fw.write(\"\\n\")\n",
    "    fw.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gender vs country\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India', 'Indonesia']\n",
      "removed categories : ['Canada']\n",
      "male : India > UK > France > Germany > USA > Italy > Brazil > Mexico = Spain > Indonesia\n",
      "female : Indonesia > Spain = Mexico > Brazil > Italy > USA > Germany > France > UK > India\n",
      "\n",
      "age vs country\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India', 'Indonesia']\n",
      "removed categories : ['Canada']\n",
      "<15y : Germany > USA > UK = Italy > Spain > France > Indonesia > Brazil > Mexico > India\n",
      "15y-25y : Brazil > USA > Germany > France > UK > Italy > Indonesia = Spain > Mexico > India\n",
      "25y-35y : India > Indonesia > Mexico > Spain > Italy > France > UK > Germany = Brazil > USA\n",
      "35y-45y : Mexico > India > UK > Spain > France > Germany > Italy = USA > Brazil > Indonesia\n",
      "45-75y : Germany > UK = France > Spain = Italy > Mexico > USA > India > Brazil > Indonesia\n",
      "\n",
      "skin tone vs country\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India', 'Indonesia']\n",
      "removed categories : ['Canada']\n",
      "light : Mexico = India > Indonesia > Spain > Italy > Germany > France > Brazil > UK > USA\n",
      "medium : USA > UK > Brazil > Germany > France > Italy > Spain > Indonesia > India > Mexico\n",
      "dark : USA > UK > France > Brazil > Italy > Germany > Spain > Indonesia > Mexico > India\n",
      "\n",
      "gender vs brand\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KERASTASE', 'LOP', 'KIEHLS', 'NYX', 'PRADA', 'REDKEN', 'BIOTHERM']\n",
      "removed categories : []\n",
      "male : PRADA > ARMANI > LRP > VICHY > OAP > YSL > CERAVE > KIEHLS > NYX > KERASTASE > GARNIER > BIOTHERM > LOP > LANCOME > REDKEN > MNY\n",
      "female : MNY > REDKEN > LANCOME > LOP > BIOTHERM > GARNIER > KERASTASE > NYX > KIEHLS > CERAVE > YSL > OAP > VICHY > LRP > ARMANI > PRADA\n",
      "\n",
      "age vs brand\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KERASTASE', 'LOP', 'KIEHLS', 'NYX', 'PRADA', 'REDKEN', 'BIOTHERM']\n",
      "removed categories : []\n",
      "<15y : CERAVE > GARNIER > NYX > LOP > MNY > LRP > YSL > KERASTASE > PRADA > KIEHLS > REDKEN > VICHY > ARMANI > BIOTHERM > OAP > LANCOME\n",
      "15y-25y : LOP > REDKEN > MNY > NYX > YSL > KERASTASE > LRP > GARNIER > BIOTHERM > VICHY > CERAVE > LANCOME > ARMANI > KIEHLS > OAP > PRADA\n",
      "25y-35y : PRADA > ARMANI > OAP > LANCOME > BIOTHERM > CERAVE > VICHY > KIEHLS > GARNIER > LRP > KERASTASE > YSL > NYX > MNY > LOP > REDKEN\n",
      "35y-45y : REDKEN > KIEHLS > OAP > VICHY > LANCOME > BIOTHERM > KERASTASE > ARMANI > CERAVE > PRADA > LRP > NYX > YSL > MNY > LOP > GARNIER\n",
      "45-75y : KIEHLS > VICHY > OAP > KERASTASE > ARMANI > YSL > BIOTHERM > LRP > LANCOME > MNY > NYX > GARNIER > CERAVE > LOP > PRADA > REDKEN\n",
      "\n",
      "skin tone vs brand\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KERASTASE', 'LOP', 'KIEHLS', 'NYX', 'PRADA', 'REDKEN', 'BIOTHERM']\n",
      "removed categories : []\n",
      "light : LOP > CERAVE > KERASTASE > BIOTHERM > VICHY > PRADA > REDKEN > GARNIER > LANCOME > OAP > MNY > LRP > KIEHLS > ARMANI > YSL > NYX\n",
      "medium : KIEHLS > NYX > MNY > REDKEN > OAP > ARMANI > LRP > LANCOME > GARNIER > YSL > VICHY > PRADA > LOP > CERAVE > KERASTASE > BIOTHERM\n",
      "dark : YSL > NYX > ARMANI > LRP > PRADA > BIOTHERM > KIEHLS > OAP > MNY > GARNIER > KERASTASE > LANCOME > VICHY > CERAVE > LOP > REDKEN\n",
      "\n",
      "gender vs product type\n",
      "product type categories : ['fragrance', 'haircare', 'makeup', 'skincare']\n",
      "removed categories : []\n",
      "male : fragrance > skincare > makeup > haircare\n",
      "female : haircare > makeup > skincare > fragrance\n",
      "\n",
      "age vs product type\n",
      "product type categories : ['fragrance', 'haircare', 'makeup', 'skincare']\n",
      "removed categories : []\n",
      "<15y : makeup = haircare > skincare > fragrance\n",
      "15y-25y : makeup > haircare > skincare > fragrance\n",
      "25y-35y : fragrance > skincare > haircare > makeup\n",
      "35y-45y : skincare > fragrance > makeup > haircare\n",
      "45-75y : skincare > makeup > fragrance > haircare\n",
      "\n",
      "skin tone vs product type\n",
      "product type categories : ['fragrance', 'haircare', 'makeup', 'skincare']\n",
      "removed categories : []\n",
      "light : haircare > fragrance > skincare > makeup\n",
      "medium : makeup > skincare > haircare > fragrance\n",
      "dark : fragrance > makeup > skincare > haircare\n",
      "\n",
      "gender vs product target gender\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "male : male > unisex > female > unspecified\n",
      "female : unspecified > female > unisex > male\n",
      "\n",
      "age vs product target gender\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "<15y : unisex > unspecified > female > male\n",
      "15y-25y : unspecified > unisex = female > male\n",
      "25y-35y : male > female > unisex = unspecified\n",
      "35y-45y : male > female > unisex = unspecified\n",
      "45-75y : female > unspecified > unisex = male\n",
      "\n",
      "skin tone vs product target gender\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "light : unisex > unspecified = female = male\n",
      "medium : male = unspecified > female > unisex\n",
      "dark : female = male > unspecified > unisex\n",
      "\n"
     ]
    }
   ],
   "source": [
    "metadata_vars = [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "rp_vars = [\"gender\", \"age\", \"skin tone\"]\n",
    "\n",
    "for metadata in metadata_vars:\n",
    "    for rp in rp_vars:\n",
    "        metadata_x_rp(data_df, rp_data2, metadata, rp)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Demographic Identity vs Metadata vs Metadata BarPlots"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "country x brand\n",
      "52 (country, brand) pairs have atleast 30 samples\n",
      "6 countrys have atleast 2 brands with atleast 30 samples: ['France' 'Germany' 'Italy' 'Spain' 'UK' 'USA']\n",
      "13 brands have atleast 2 countrys with atleast 30 samples: ['ARMANI' 'CERAVE' 'GARNIER' 'KERASTASE' 'KIEHLS' 'LANCOME' 'LOP' 'LRP'\n",
      " 'MAYBELLINE' 'NYX' 'OAP' 'VICHY' 'YSL']\n",
      "\n",
      "country x product type\n",
      "26 (country, product type) pairs have atleast 30 samples\n",
      "6 countrys have atleast 2 product types with atleast 30 samples: ['France' 'Germany' 'Italy' 'Spain' 'UK' 'USA']\n",
      "4 product types have atleast 2 countrys with atleast 30 samples: ['fragrance' 'haircare' 'makeup' 'skincare']\n",
      "\n",
      "country x product target gender\n",
      "20 (country, product target gender) pairs have atleast 30 samples\n",
      "6 countrys have atleast 2 product target genders with atleast 30 samples: ['France' 'Germany' 'Italy' 'Spain' 'UK' 'USA']\n",
      "4 product target genders have atleast 2 countrys with atleast 30 samples: ['female' 'male' 'unisex' 'unspecified']\n",
      "\n",
      "brand x product type\n",
      "30 (brand, product type) pairs have atleast 30 samples\n",
      "9 brands have atleast 2 product types with atleast 30 samples: ['ARMANI' 'GARNIER' 'KIEHLS' 'LANCOME' 'LRP' 'MNY' 'OAP' 'VICHY' 'YSL']\n",
      "4 product types have atleast 2 brands with atleast 30 samples: ['fragrance' 'haircare' 'makeup' 'skincare']\n",
      "\n",
      "brand x product target gender\n",
      "29 (brand, product target gender) pairs have atleast 30 samples\n",
      "9 brands have atleast 2 product target genders with atleast 30 samples: ['ARMANI' 'CERAVE' 'GARNIER' 'KIEHLS' 'LANCOME' 'LRP' 'OAP' 'VICHY' 'YSL']\n",
      "4 product target genders have atleast 2 brands with atleast 30 samples: ['female' 'male' 'unisex' 'unspecified']\n",
      "\n",
      "product type x product target gender\n",
      "10 (product type, product target gender) pairs have atleast 30 samples\n",
      "3 product types have atleast 2 product target genders with atleast 30 samples: ['fragrance' 'haircare' 'skincare']\n",
      "3 product target genders have atleast 2 product types with atleast 30 samples: ['female' 'male' 'unspecified']\n",
      "\n"
     ]
    }
   ],
   "source": [
    "metadata_columns = [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "\n",
    "for i in range(len(metadata_columns)):\n",
    "    for j in range(i + 1, len(metadata_columns)):\n",
    "        col1 = metadata_columns[i]\n",
    "        col2 = metadata_columns[j]\n",
    "        gdf = data_df.groupby([col1, col2]).count()[\"Creative ID\"].reset_index().rename(columns={\"Creative ID\": \"n\"})\n",
    "        print(f\"{col1} x {col2}\")\n",
    "        n = (gdf.n >= 30).sum()\n",
    "        gdf1 = gdf[gdf.n >= 30].groupby(col1).count()[\"n\"].reset_index()\n",
    "        n1 = (gdf1.n > 1).sum()\n",
    "        cats1 = gdf1.loc[gdf1.n > 1, col1].values\n",
    "        gdf2 = gdf[gdf.n >= 30].groupby(col2).count()[\"n\"].reset_index()\n",
    "        n2 = (gdf2.n > 1).sum()\n",
    "        cats2 = gdf2.loc[gdf2.n > 1, col2].values\n",
    "        print(f\"{n} ({col1}, {col2}) pairs have atleast 30 samples\")\n",
    "        print(f\"{n1} {col1}s have atleast 2 {col2}s with atleast 30 samples: {cats1}\")\n",
    "        print(f\"{n2} {col2}s have atleast 2 {col1}s with atleast 30 samples: {cats2}\")\n",
    "        print()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def metadata2_x_rp(df, rp_data, metadata1, metadata2, rp, confidence=0.05):\n",
    "    assert rp in [\"gender\", \"age\", \"skin tone\"]\n",
    "    assert metadata1 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    assert metadata2 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    metadata1_cats = df[metadata1].unique()\n",
    "    metadata2_cats = df[metadata2].unique()\n",
    "    \n",
    "    for metadata1_cat in metadata1_cats:\n",
    "        index = df[metadata1] == metadata1_cat\n",
    "        fdf = df[index].copy()\n",
    "        fdf.index = pd.RangeIndex(0, len(fdf))\n",
    "        metadata_x_rp(fdf, rp_data[index], metadata=metadata2, rp=rp, confidence=confidence, \n",
    "                      condition=f\"{metadata1}={metadata1_cat}\")\n",
    "    \n",
    "    for metadata2_cat in metadata2_cats:\n",
    "        index = df[metadata2] == metadata2_cat\n",
    "        fdf = df[index].copy()\n",
    "        fdf.index = pd.RangeIndex(0, len(fdf))\n",
    "        metadata_x_rp(fdf, rp_data[index], metadata=metadata1, rp=rp, confidence=confidence, \n",
    "                      condition=f\"{metadata2}={metadata2_cat}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "gender vs brand country=USA\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'MNY', 'NYX', 'OAP', 'YSL']\n",
      "removed categories : ['KIEHLS', 'LANCOME', 'LRP', 'PRADA', 'REDKEN', 'VICHY']\n",
      "male : CERAVE = YSL > OAP > ARMANI = MNY > GARNIER > NYX\n",
      "female : NYX > GARNIER > MNY = ARMANI > OAP > YSL = CERAVE\n",
      "\n",
      "gender vs brand country=France\n",
      "brand categories : ['GARNIER', 'KIEHLS', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'NYX', 'PRADA', 'YSL']\n",
      "male : KIEHLS > LRP = VICHY > GARNIER > LANCOME > OAP > MNY\n",
      "female : MNY > OAP > LANCOME > GARNIER > VICHY = LRP > KIEHLS\n",
      "\n",
      "gender vs brand country=Germany\n",
      "brand categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'VICHY', 'YSL']\n",
      "removed categories : ['PRADA', 'REDKEN']\n",
      "male : VICHY > LRP > ARMANI > NYX > YSL > CERAVE > KERASTASE > GARNIER > KIEHLS > BIOTHERM > LANCOME > MNY > LOP\n",
      "female : LOP > MNY > LANCOME > BIOTHERM > KIEHLS > GARNIER > KERASTASE > CERAVE > YSL > NYX > ARMANI > LRP > VICHY\n",
      "\n",
      "gender vs brand country=Italy\n",
      "brand categories : ['ARMANI', 'GARNIER', 'KIEHLS', 'LANCOME', 'MNY', 'OAP', 'YSL']\n",
      "removed categories : ['BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'LRP', 'NYX', 'PRADA', 'REDKEN']\n",
      "male : ARMANI > YSL = LANCOME > KIEHLS > OAP = GARNIER > MNY\n",
      "female : MNY > GARNIER = OAP > KIEHLS > LANCOME = YSL > ARMANI\n",
      "\n",
      "gender vs brand country=Spain\n",
      "brand categories : ['GARNIER', 'LANCOME', 'OAP']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'KERASTASE', 'KIEHLS', 'LOP', 'LRP', 'MNY', 'NYX', 'REDKEN', 'VICHY', 'YSL']\n",
      "male : GARNIER = OAP = LANCOME\n",
      "female : LANCOME = OAP = GARNIER\n",
      "\n",
      "gender vs brand country=UK\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'OAP', 'PRADA', 'VICHY', 'YSL']\n",
      "removed categories : ['REDKEN']\n",
      "male : PRADA > OAP > LRP > YSL > ARMANI > VICHY > CERAVE > LOP > NYX > KIEHLS > GARNIER > MNY > KERASTASE > LANCOME\n",
      "female : LANCOME > KERASTASE > MNY > GARNIER > KIEHLS > NYX > LOP > CERAVE > VICHY > ARMANI > YSL > LRP > OAP > PRADA\n",
      "\n",
      "gender vs country brand=ARMANI\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'France', 'Spain']\n",
      "male : Italy > Germany > UK > USA\n",
      "female : USA > UK > Germany > Italy\n",
      "\n",
      "gender vs country brand=GARNIER\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "male : India > Germany = France > UK > USA > Spain = Italy\n",
      "female : Italy = Spain > USA > UK > France = Germany > India\n",
      "\n",
      "gender vs country brand=LANCOME\n",
      "country categories : ['France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Canada']\n",
      "male : Italy = Germany = France > Spain = UK\n",
      "female : UK = Spain > France = Germany = Italy\n",
      "\n",
      "gender vs country brand=LRP\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Italy', 'Spain']\n",
      "male : UK = France = Germany\n",
      "female : Germany = France = UK\n",
      "\n",
      "gender vs country brand=MNY\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Spain', 'Indonesia']\n",
      "male : UK = USA > Germany > France = Italy\n",
      "female : Italy = France > Germany > USA = UK\n",
      "\n",
      "gender vs country brand=OAP\n",
      "country categories : ['USA', 'France', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'India']\n",
      "male : UK > USA > France > Spain = Italy\n",
      "female : Italy = Spain > France > USA > UK\n",
      "\n",
      "gender vs country brand=VICHY\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'Spain']\n",
      "male : Germany = France > UK\n",
      "female : UK > France = Germany\n",
      "\n",
      "gender vs country brand=YSL\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'France', 'Spain']\n",
      "male : UK = Germany = USA = Italy\n",
      "female : Italy = USA = Germany = UK\n",
      "\n",
      "gender vs country brand=CERAVE\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy']\n",
      "male : USA = Germany = UK\n",
      "female : UK = Germany = USA\n",
      "\n",
      "gender vs country brand=KERASTASE\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "male : Germany > UK\n",
      "female : UK > Germany\n",
      "\n",
      "gender vs country brand=LOP\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "male : UK > Germany\n",
      "female : Germany > UK\n",
      "\n",
      "gender vs country brand=KIEHLS\n",
      "country categories : ['France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['USA', 'Spain']\n",
      "male : France > Germany > UK = Italy\n",
      "female : Italy = UK > Germany > France\n",
      "\n",
      "gender vs country brand=NYX\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['France', 'Italy', 'Spain']\n",
      "male : Germany > UK > USA\n",
      "female : USA > UK > Germany\n",
      "\n",
      "age vs brand country=USA\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'MNY', 'NYX', 'OAP', 'YSL']\n",
      "removed categories : ['KIEHLS', 'LANCOME', 'LRP', 'PRADA', 'REDKEN', 'VICHY']\n",
      "<15y : MNY > CERAVE > YSL > NYX = ARMANI = GARNIER > OAP\n",
      "15y-25y : NYX > ARMANI > MNY = YSL > GARNIER > CERAVE > OAP\n",
      "25y-35y : OAP > CERAVE > GARNIER > YSL > MNY = ARMANI = NYX\n",
      "35y-45y : OAP > MNY = CERAVE = GARNIER > YSL > ARMANI > NYX\n",
      "45-75y : ARMANI > OAP > GARNIER = YSL > CERAVE > MNY > NYX\n",
      "\n",
      "age vs brand country=France\n",
      "brand categories : ['GARNIER', 'KIEHLS', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'NYX', 'PRADA', 'YSL']\n",
      "<15y : GARNIER > MNY = VICHY > KIEHLS = LRP > OAP = LANCOME\n",
      "15y-25y : LRP > VICHY > MNY > OAP = GARNIER > LANCOME = KIEHLS\n",
      "25y-35y : KIEHLS > LANCOME > OAP > GARNIER = MNY > VICHY > LRP\n",
      "35y-45y : OAP > GARNIER > VICHY = LANCOME > LRP = MNY > KIEHLS\n",
      "45-75y : OAP > VICHY > LRP > GARNIER > LANCOME = MNY > KIEHLS\n",
      "\n",
      "age vs brand country=Germany\n",
      "brand categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'VICHY', 'YSL']\n",
      "removed categories : ['PRADA', 'REDKEN']\n",
      "<15y : GARNIER > LRP > YSL > MNY > NYX > KERASTASE > KIEHLS > LOP > CERAVE > ARMANI > VICHY > BIOTHERM > LANCOME\n",
      "15y-25y : LOP > MNY > NYX > YSL > GARNIER > KERASTASE > LRP > BIOTHERM > CERAVE > VICHY > LANCOME > KIEHLS > ARMANI\n",
      "25y-35y : ARMANI > LANCOME > CERAVE > BIOTHERM > VICHY > LRP > KIEHLS > KERASTASE > NYX > GARNIER > YSL > LOP > MNY\n",
      "35y-45y : KIEHLS > VICHY > LANCOME > YSL > BIOTHERM > KERASTASE > ARMANI > LOP > CERAVE > LRP > MNY > GARNIER > NYX\n",
      "45-75y : KIEHLS > VICHY > KERASTASE > ARMANI > YSL > BIOTHERM > LRP > LANCOME > MNY > LOP > CERAVE > GARNIER > NYX\n",
      "\n",
      "age vs brand country=Italy\n",
      "brand categories : ['ARMANI', 'GARNIER', 'KIEHLS', 'LANCOME', 'MNY', 'OAP', 'YSL']\n",
      "removed categories : ['BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'LRP', 'NYX', 'PRADA', 'REDKEN']\n",
      "<15y : GARNIER > MNY > YSL = OAP > ARMANI > LANCOME > KIEHLS\n",
      "15y-25y : MNY > YSL > LANCOME = GARNIER = OAP > KIEHLS > ARMANI\n",
      "25y-35y : ARMANI > OAP > LANCOME > GARNIER > YSL = KIEHLS > MNY\n",
      "35y-45y : KIEHLS = LANCOME > OAP > YSL = ARMANI > GARNIER > MNY\n",
      "45-75y : KIEHLS > ARMANI > LANCOME = OAP > GARNIER > MNY > YSL\n",
      "\n",
      "age vs brand country=Spain\n",
      "brand categories : ['GARNIER', 'LANCOME', 'OAP']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'KERASTASE', 'KIEHLS', 'LOP', 'LRP', 'MNY', 'NYX', 'REDKEN', 'VICHY', 'YSL']\n",
      "<15y : GARNIER > OAP = LANCOME\n",
      "15y-25y : GARNIER = LANCOME = OAP\n",
      "25y-35y : OAP = LANCOME > GARNIER\n",
      "35y-45y : LANCOME = OAP = GARNIER\n",
      "45-75y : OAP = GARNIER = LANCOME\n",
      "\n",
      "age vs brand country=UK\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'OAP', 'PRADA', 'VICHY', 'YSL']\n",
      "removed categories : ['REDKEN']\n",
      "<15y : CERAVE > LOP > NYX > MNY > GARNIER > KERASTASE > KIEHLS > YSL > ARMANI > OAP > PRADA > LANCOME > LRP > VICHY\n",
      "15y-25y : LOP > LRP > KERASTASE > NYX > YSL > MNY > ARMANI > LANCOME > GARNIER > KIEHLS > VICHY > CERAVE > OAP > PRADA\n",
      "25y-35y : PRADA > OAP > KIEHLS > GARNIER > LANCOME > MNY > ARMANI > YSL > VICHY > KERASTASE > NYX > CERAVE > LRP > LOP\n",
      "35y-45y : OAP > VICHY > ARMANI > KIEHLS > NYX > PRADA > LRP > YSL > GARNIER > LANCOME > CERAVE > MNY > KERASTASE > LOP\n",
      "45-75y : VICHY > KIEHLS > YSL > LRP > NYX > OAP > GARNIER > MNY > ARMANI > KERASTASE > CERAVE > LANCOME > PRADA = LOP\n",
      "\n",
      "age vs country brand=ARMANI\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'France', 'Spain']\n",
      "<15y : USA = Germany = UK > Italy\n",
      "15y-25y : USA > UK > Germany = Italy\n",
      "25y-35y : Italy > Germany > UK > USA\n",
      "35y-45y : UK > Germany > Italy > USA\n",
      "45-75y : USA = Italy = Germany = UK\n",
      "\n",
      "age vs country brand=GARNIER\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "<15y : Italy > France > Germany = Spain > UK > USA > India\n",
      "15y-25y : Germany = USA > UK > Spain = France > Italy > India\n",
      "25y-35y : India > Spain = UK = Italy > France > USA > Germany\n",
      "35y-45y : India > France > Spain > USA = UK > Italy > Germany\n",
      "45-75y : Spain = UK = France = USA > Italy = India > Germany\n",
      "\n",
      "age vs country brand=LANCOME\n",
      "country categories : ['France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Canada']\n",
      "<15y : France = Spain = Germany = UK = Italy\n",
      "15y-25y : UK > Italy = France = Germany = Spain\n",
      "25y-35y : Spain = Germany = France = Italy > UK\n",
      "35y-45y : Spain = Italy = Germany > France > UK\n",
      "45-75y : Italy = Germany = France = Spain > UK\n",
      "\n",
      "age vs country brand=LRP\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Italy', 'Spain']\n",
      "<15y : Germany > France > UK\n",
      "15y-25y : UK = France > Germany\n",
      "25y-35y : Germany > France = UK\n",
      "35y-45y : UK = France = Germany\n",
      "45-75y : France = UK > Germany\n",
      "\n",
      "age vs country brand=MNY\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Spain', 'Indonesia']\n",
      "<15y : USA > Germany = Italy = UK > France\n",
      "15y-25y : Germany = USA = Italy > UK = France\n",
      "25y-35y : France > UK > Italy > Germany = USA\n",
      "35y-45y : USA = Germany = France > UK > Italy\n",
      "45-75y : UK = Germany > France > USA = Italy\n",
      "\n",
      "age vs country brand=OAP\n",
      "country categories : ['USA', 'France', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'India']\n",
      "<15y : USA = Spain = UK = France = Italy\n",
      "15y-25y : France > Italy = USA > Spain > UK\n",
      "25y-35y : Spain = UK > Italy > USA > France\n",
      "35y-45y : UK = USA > France > Spain > Italy\n",
      "45-75y : France = Spain > UK = USA = Italy\n",
      "\n",
      "age vs country brand=VICHY\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'Spain']\n",
      "<15y : France > Germany > UK\n",
      "15y-25y : France > Germany = UK\n",
      "25y-35y : Germany > UK = France\n",
      "35y-45y : UK > Germany > France\n",
      "45-75y : UK > Germany > France\n",
      "\n",
      "age vs country brand=YSL\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'France', 'Spain']\n",
      "<15y : Germany = USA > UK > Italy\n",
      "15y-25y : USA > UK = Germany = Italy\n",
      "25y-35y : Italy > UK = Germany = USA\n",
      "35y-45y : Germany > UK > Italy = USA\n",
      "45-75y : UK > Germany = USA > Italy\n",
      "\n",
      "age vs country brand=CERAVE\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy']\n",
      "<15y : UK > USA > Germany\n",
      "15y-25y : USA = Germany > UK\n",
      "25y-35y : Germany > USA > UK\n",
      "35y-45y : USA = Germany = UK\n",
      "45-75y : UK = USA = Germany\n",
      "\n",
      "age vs country brand=KERASTASE\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "<15y : UK = Germany\n",
      "15y-25y : UK > Germany\n",
      "25y-35y : Germany = UK\n",
      "35y-45y : Germany > UK\n",
      "45-75y : Germany > UK\n",
      "\n",
      "age vs country brand=LOP\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "<15y : UK > Germany\n",
      "15y-25y : UK = Germany\n",
      "25y-35y : Germany > UK\n",
      "35y-45y : Germany > UK\n",
      "45-75y : Germany > UK\n",
      "\n",
      "age vs country brand=KIEHLS\n",
      "country categories : ['France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['USA', 'Spain']\n",
      "<15y : UK = Germany = France > Italy\n",
      "15y-25y : UK > France = Italy = Germany\n",
      "25y-35y : France > UK = Italy = Germany\n",
      "35y-45y : Germany > Italy = UK > France\n",
      "45-75y : Italy = Germany > UK > France\n",
      "\n",
      "age vs country brand=NYX\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['France', 'Italy', 'Spain']\n",
      "<15y : UK > Germany > USA\n",
      "15y-25y : USA > Germany = UK\n",
      "25y-35y : Germany > UK = USA\n",
      "35y-45y : UK > Germany > USA\n",
      "45-75y : UK > USA = Germany\n",
      "\n",
      "skin tone vs brand country=USA\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'MNY', 'NYX', 'OAP', 'YSL']\n",
      "removed categories : ['KIEHLS', 'LANCOME', 'LRP', 'PRADA', 'REDKEN', 'VICHY']\n",
      "light : CERAVE > YSL = GARNIER = OAP > MNY = NYX > ARMANI\n",
      "medium : ARMANI > MNY = NYX > GARNIER = OAP > YSL = CERAVE\n",
      "dark : ARMANI = YSL > OAP > NYX = GARNIER > MNY > CERAVE\n",
      "\n",
      "skin tone vs brand country=France\n",
      "brand categories : ['GARNIER', 'KIEHLS', 'LANCOME', 'LRP', 'MNY', 'OAP', 'VICHY']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'NYX', 'PRADA', 'YSL']\n",
      "light : GARNIER = VICHY > LANCOME > LRP > OAP = MNY > KIEHLS\n",
      "medium : KIEHLS > MNY = OAP > LANCOME = LRP > GARNIER > VICHY\n",
      "dark : KIEHLS = LRP = VICHY = MNY = OAP > LANCOME = GARNIER\n",
      "\n",
      "skin tone vs brand country=Germany\n",
      "brand categories : ['ARMANI', 'BIOTHERM', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'VICHY', 'YSL']\n",
      "removed categories : ['PRADA', 'REDKEN']\n",
      "light : CERAVE > KERASTASE > LOP > VICHY > BIOTHERM > GARNIER > LANCOME > MNY > KIEHLS > YSL > ARMANI > NYX > LRP\n",
      "medium : KIEHLS > NYX > LRP > MNY > LANCOME > YSL > GARNIER > VICHY > ARMANI > LOP > CERAVE > BIOTHERM > KERASTASE\n",
      "dark : LRP > ARMANI > NYX > YSL > BIOTHERM > LANCOME > KERASTASE > GARNIER > MNY > KIEHLS > CERAVE > VICHY > LOP\n",
      "\n",
      "skin tone vs brand country=Italy\n",
      "brand categories : ['ARMANI', 'GARNIER', 'KIEHLS', 'LANCOME', 'MNY', 'OAP', 'YSL']\n",
      "removed categories : ['BIOTHERM', 'CERAVE', 'KERASTASE', 'LOP', 'LRP', 'NYX', 'PRADA', 'REDKEN']\n",
      "light : GARNIER > OAP = KIEHLS > MNY = ARMANI = LANCOME > YSL\n",
      "medium : KIEHLS = YSL > ARMANI = MNY = LANCOME > OAP > GARNIER\n",
      "dark : YSL > LANCOME > ARMANI = MNY > OAP > GARNIER > KIEHLS\n",
      "\n",
      "skin tone vs brand country=Spain\n",
      "brand categories : ['GARNIER', 'LANCOME', 'OAP']\n",
      "removed categories : ['ARMANI', 'BIOTHERM', 'KERASTASE', 'KIEHLS', 'LOP', 'LRP', 'MNY', 'NYX', 'REDKEN', 'VICHY', 'YSL']\n",
      "light : LANCOME = OAP = GARNIER\n",
      "medium : GARNIER = OAP = LANCOME\n",
      "dark : LANCOME = OAP = GARNIER\n",
      "\n",
      "skin tone vs brand country=UK\n",
      "brand categories : ['ARMANI', 'CERAVE', 'GARNIER', 'KERASTASE', 'KIEHLS', 'LANCOME', 'LOP', 'LRP', 'MNY', 'NYX', 'OAP', 'PRADA', 'VICHY', 'YSL']\n",
      "removed categories : ['REDKEN']\n",
      "light : LOP > CERAVE > PRADA > LRP > OAP > KERASTASE > GARNIER > LANCOME > YSL > VICHY > ARMANI > MNY > KIEHLS > NYX\n",
      "medium : MNY > VICHY > KIEHLS > LANCOME > NYX > OAP > GARNIER > ARMANI > KERASTASE > LRP > YSL > CERAVE > PRADA > LOP\n",
      "dark : NYX > ARMANI > YSL > KIEHLS > MNY > GARNIER > KERASTASE > VICHY > PRADA > LRP > OAP > LANCOME > CERAVE > LOP\n",
      "\n",
      "skin tone vs country brand=ARMANI\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'France', 'Spain']\n",
      "light : Italy > Germany > UK > USA\n",
      "medium : USA > UK > Italy = Germany\n",
      "dark : UK = USA = Germany > Italy\n",
      "\n",
      "skin tone vs country brand=GARNIER\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "light : India > Italy > Spain = France > Germany > UK = USA\n",
      "medium : USA > UK > Germany = Spain > France = Italy > India\n",
      "dark : UK > USA > France > Germany = Italy > Spain > India\n",
      "\n",
      "skin tone vs country brand=LANCOME\n",
      "country categories : ['France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Canada']\n",
      "light : Spain > France = Germany = Italy > UK\n",
      "medium : UK > Germany = France = Italy > Spain\n",
      "dark : Italy > France = Germany = UK = Spain\n",
      "\n",
      "skin tone vs country brand=LRP\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'USA', 'Italy', 'Spain']\n",
      "light : France = UK > Germany\n",
      "medium : Germany = UK = France\n",
      "dark : Germany > France = UK\n",
      "\n",
      "skin tone vs country brand=MNY\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Spain', 'Indonesia']\n",
      "light : Italy = Germany > France > USA > UK\n",
      "medium : UK = USA > France = Germany > Italy\n",
      "dark : UK > USA = France = Italy > Germany\n",
      "\n",
      "skin tone vs country brand=OAP\n",
      "country categories : ['USA', 'France', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'India']\n",
      "light : Spain > Italy > France = UK > USA\n",
      "medium : USA = UK > France > Italy = Spain\n",
      "dark : USA > France > Italy = UK > Spain\n",
      "\n",
      "skin tone vs country brand=VICHY\n",
      "country categories : ['France', 'Germany', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'Spain']\n",
      "light : Germany = France > UK\n",
      "medium : UK > Germany > France\n",
      "dark : UK = France > Germany\n",
      "\n",
      "skin tone vs country brand=YSL\n",
      "country categories : ['USA', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'France', 'Spain']\n",
      "light : Germany > USA = Italy = UK\n",
      "medium : Germany = Italy = UK = USA\n",
      "dark : UK = USA = Italy > Germany\n",
      "\n",
      "skin tone vs country brand=CERAVE\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy']\n",
      "light : Germany > UK = USA\n",
      "medium : USA = UK = Germany\n",
      "dark : USA = UK > Germany\n",
      "\n",
      "skin tone vs country brand=KERASTASE\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "light : Germany > UK\n",
      "medium : UK > Germany\n",
      "dark : UK > Germany\n",
      "\n",
      "skin tone vs country brand=LOP\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'France', 'Italy', 'Spain']\n",
      "light : UK = Germany\n",
      "medium : Germany = UK\n",
      "dark : Germany = UK\n",
      "\n",
      "skin tone vs country brand=KIEHLS\n",
      "country categories : ['France', 'Germany', 'Italy', 'UK']\n",
      "removed categories : ['USA', 'Spain']\n",
      "light : Italy > Germany > France > UK\n",
      "medium : UK > Germany = France > Italy\n",
      "dark : UK > France > Germany > Italy\n",
      "\n",
      "skin tone vs country brand=NYX\n",
      "country categories : ['USA', 'Germany', 'UK']\n",
      "removed categories : ['France', 'Italy', 'Spain']\n",
      "light : Germany = USA = UK\n",
      "medium : USA = UK = Germany\n",
      "dark : UK > USA = Germany\n",
      "\n",
      "gender vs product type country=USA\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "male : skincare > makeup = fragrance > haircare\n",
      "female : haircare > fragrance = makeup > skincare\n",
      "\n",
      "gender vs product type country=France\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "male : skincare > fragrance > haircare = makeup\n",
      "female : makeup = haircare > fragrance > skincare\n",
      "\n",
      "gender vs product type country=Germany\n",
      "product type categories : ['skincare', 'fragrance', 'makeup', 'haircare']\n",
      "removed categories : []\n",
      "male : fragrance > makeup = skincare > haircare\n",
      "female : haircare > skincare = makeup > fragrance\n",
      "\n",
      "gender vs product type country=Italy\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "male : fragrance > makeup > haircare = skincare\n",
      "female : skincare = haircare > makeup > fragrance\n",
      "\n",
      "gender vs product type country=Spain\n",
      "product type categories : ['fragrance', 'skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "male : fragrance > makeup = skincare = haircare\n",
      "female : haircare = skincare = makeup > fragrance\n",
      "\n",
      "gender vs product type country=UK\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "male : fragrance > skincare > makeup > haircare\n",
      "female : haircare > makeup > skincare > fragrance\n",
      "\n",
      "gender vs country product type=fragrance\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil']\n",
      "male : Italy > UK > Spain > Germany > France = USA\n",
      "female : USA = France > Germany > Spain > UK > Italy\n",
      "\n",
      "gender vs country product type=haircare\n",
      "country categories : ['Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "male : India > Germany > France > Brazil = UK > USA = Italy > Spain\n",
      "female : Spain > Italy = USA > UK = Brazil > France > Germany > India\n",
      "\n",
      "gender vs country product type=makeup\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Indonesia']\n",
      "male : Germany > USA > UK > France > Italy > Spain\n",
      "female : Spain > Italy > France > UK > USA > Germany\n",
      "\n",
      "gender vs country product type=skincare\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'India', 'Indonesia']\n",
      "male : UK > France > USA > Germany > Spain = Italy\n",
      "female : Italy = Spain > Germany > USA > France > UK\n",
      "\n",
      "age vs product type country=USA\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "<15y : makeup = skincare = fragrance > haircare\n",
      "15y-25y : fragrance = makeup > haircare > skincare\n",
      "25y-35y : skincare = haircare > makeup = fragrance\n",
      "35y-45y : skincare > haircare = makeup = fragrance\n",
      "45-75y : fragrance = haircare = skincare = makeup\n",
      "\n",
      "age vs product type country=France\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "<15y : makeup = haircare > skincare > fragrance\n",
      "15y-25y : skincare > makeup > haircare > fragrance\n",
      "25y-35y : fragrance > haircare > makeup > skincare\n",
      "35y-45y : haircare = fragrance > skincare = makeup\n",
      "45-75y : skincare = makeup > haircare > fragrance\n",
      "\n",
      "age vs product type country=Germany\n",
      "product type categories : ['skincare', 'fragrance', 'makeup', 'haircare']\n",
      "removed categories : []\n",
      "<15y : haircare = makeup > skincare > fragrance\n",
      "15y-25y : makeup = haircare > skincare > fragrance\n",
      "25y-35y : fragrance > skincare > haircare > makeup\n",
      "35y-45y : fragrance = makeup = skincare > haircare\n",
      "45-75y : makeup = skincare = fragrance > haircare\n",
      "\n",
      "age vs product type country=Italy\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "<15y : haircare = makeup > fragrance = skincare\n",
      "15y-25y : makeup > haircare > skincare = fragrance\n",
      "25y-35y : fragrance > skincare > haircare > makeup\n",
      "35y-45y : skincare > haircare = fragrance > makeup\n",
      "45-75y : skincare > fragrance = makeup = haircare\n",
      "\n",
      "age vs product type country=Spain\n",
      "product type categories : ['fragrance', 'skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "<15y : haircare > makeup = skincare > fragrance\n",
      "15y-25y : makeup > haircare = skincare > fragrance\n",
      "25y-35y : fragrance > skincare = haircare > makeup\n",
      "35y-45y : fragrance = skincare > haircare = makeup\n",
      "45-75y : skincare = makeup = haircare = fragrance\n",
      "\n",
      "age vs product type country=UK\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "<15y : skincare > makeup > haircare > fragrance\n",
      "15y-25y : makeup > haircare > skincare > fragrance\n",
      "25y-35y : fragrance > skincare > haircare > makeup\n",
      "35y-45y : skincare > fragrance > makeup > haircare\n",
      "45-75y : skincare > makeup > haircare = fragrance\n",
      "\n",
      "age vs country product type=fragrance\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil']\n",
      "<15y : USA > UK = Spain > Germany > Italy > France\n",
      "15y-25y : USA > UK = Germany = France = Italy > Spain\n",
      "25y-35y : Spain > Italy = France > Germany = UK > USA\n",
      "35y-45y : Spain = UK = France = Germany > Italy > USA\n",
      "45-75y : USA = Germany = Italy > Spain > UK = France\n",
      "\n",
      "age vs country product type=haircare\n",
      "country categories : ['Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "<15y : Germany > Spain = Italy > France = UK > Brazil > India > USA\n",
      "15y-25y : Brazil > Germany > UK > USA > Italy > France = Spain > India\n",
      "25y-35y : India > Spain = France = USA > Italy > UK > Germany > Brazil\n",
      "35y-45y : India > France > Italy = Spain = USA = UK = Germany > Brazil\n",
      "45-75y : USA = France = Spain > Germany = Italy > UK > India > Brazil\n",
      "\n",
      "age vs country product type=makeup\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Indonesia']\n",
      "<15y : Germany > Italy = USA = France = UK > Spain\n",
      "15y-25y : USA > UK > Germany > Italy = France = Spain\n",
      "25y-35y : Spain = Italy = France > UK > Germany > USA\n",
      "35y-45y : Germany > France = UK = Spain = USA > Italy\n",
      "45-75y : Germany = Spain = France = UK > Italy > USA\n",
      "\n",
      "age vs country product type=skincare\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'India', 'Indonesia']\n",
      "<15y : UK > USA = Germany > Spain = France > Italy\n",
      "15y-25y : France > Germany > USA = Spain = UK > Italy\n",
      "25y-35y : Italy = Spain = USA > UK > Germany > France\n",
      "35y-45y : UK > Italy = USA = Spain > Germany > France\n",
      "45-75y : Italy > UK = Germany = France = Spain > USA\n",
      "\n",
      "skin tone vs product type country=USA\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "light : skincare > makeup > haircare = fragrance\n",
      "medium : haircare > fragrance > skincare > makeup\n",
      "dark : makeup = fragrance > haircare = skincare\n",
      "\n",
      "skin tone vs product type country=France\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "light : fragrance > skincare = makeup = haircare\n",
      "medium : haircare > makeup > skincare > fragrance\n",
      "dark : skincare > makeup = fragrance > haircare\n",
      "\n",
      "skin tone vs product type country=Germany\n",
      "product type categories : ['skincare', 'fragrance', 'makeup', 'haircare']\n",
      "removed categories : []\n",
      "light : haircare > makeup > skincare = fragrance\n",
      "medium : skincare = makeup > haircare > fragrance\n",
      "dark : fragrance > skincare > makeup > haircare\n",
      "\n",
      "skin tone vs product type country=Italy\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "light : skincare = haircare > fragrance > makeup\n",
      "medium : makeup > haircare > fragrance = skincare\n",
      "dark : fragrance > makeup > skincare > haircare\n",
      "\n",
      "skin tone vs product type country=Spain\n",
      "product type categories : ['fragrance', 'skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "light : haircare > fragrance = skincare = makeup\n",
      "medium : makeup = skincare > haircare = fragrance\n",
      "dark : fragrance > makeup = skincare > haircare\n",
      "\n",
      "skin tone vs product type country=UK\n",
      "product type categories : ['fragrance', 'makeup', 'skincare', 'haircare']\n",
      "removed categories : []\n",
      "light : fragrance > haircare > skincare > makeup\n",
      "medium : makeup > haircare > skincare > fragrance\n",
      "dark : makeup > skincare = fragrance > haircare\n",
      "\n",
      "skin tone vs country product type=fragrance\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil']\n",
      "light : France > Spain > Italy = UK > Germany > USA\n",
      "medium : USA > Germany = UK = Italy > Spain = France\n",
      "dark : USA = Germany > UK = Italy > Spain = France\n",
      "\n",
      "skin tone vs country product type=haircare\n",
      "country categories : ['Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Mexico', 'Indonesia']\n",
      "light : India > Spain > Italy > Germany > UK = France > Brazil > USA\n",
      "medium : USA > France = Brazil = UK > Germany > Italy > Spain > India\n",
      "dark : Brazil > UK > USA = France > Italy > Germany > Spain > India\n",
      "\n",
      "skin tone vs country product type=makeup\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Indonesia']\n",
      "light : Spain = Germany > Italy = France > USA > UK\n",
      "medium : UK > USA > France = Germany = Spain = Italy\n",
      "dark : USA > UK > Italy = France > Spain = Germany\n",
      "\n",
      "skin tone vs country product type=skincare\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'India', 'Indonesia']\n",
      "light : Italy > Spain > France > Germany = USA > UK\n",
      "medium : USA > UK > Germany > Spain = France > Italy\n",
      "dark : UK > France > Germany > Italy = Spain = USA\n",
      "\n",
      "gender vs product target gender country=USA\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex', 'male']\n",
      "male : female = unspecified\n",
      "female : unspecified = female\n",
      "\n",
      "gender vs product target gender country=France\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male', 'unisex']\n",
      "male : female > unspecified\n",
      "female : unspecified > female\n",
      "\n",
      "gender vs product target gender country=Germany\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "male : male > unisex > unspecified > female\n",
      "female : female > unspecified > unisex > male\n",
      "\n",
      "gender vs product target gender country=Italy\n",
      "product target gender categories : ['male', 'unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "male : male > female = unspecified\n",
      "female : unspecified = female > male\n",
      "\n",
      "gender vs product target gender country=Spain\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male']\n",
      "male : unspecified = female\n",
      "female : female = unspecified\n",
      "\n",
      "gender vs product target gender country=UK\n",
      "product target gender categories : ['female', 'unspecified', 'male', 'unisex']\n",
      "removed categories : []\n",
      "male : male > unisex = female > unspecified\n",
      "female : unspecified > female = unisex > male\n",
      "\n",
      "gender vs country product target gender=female\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'Indonesia', 'India']\n",
      "male : France > USA = UK > Germany > Italy > Spain\n",
      "female : Spain > Italy > Germany > UK = USA > France\n",
      "\n",
      "gender vs country product target gender=male\n",
      "country categories : ['Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'France', 'Spain', 'India', 'Indonesia']\n",
      "male : UK = Italy > Germany\n",
      "female : Germany > Italy = UK\n",
      "\n",
      "gender vs country product target gender=unspecified\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Indonesia']\n",
      "male : Germany > USA > India = France > Brazil > UK > Italy > Spain > Mexico\n",
      "female : Mexico > Spain > Italy > UK > Brazil > France = India > USA > Germany\n",
      "\n",
      "gender vs country product target gender=unisex\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'USA', 'France', 'Italy']\n",
      "male : Germany = UK\n",
      "female : UK = Germany\n",
      "\n",
      "age vs product target gender country=USA\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex', 'male']\n",
      "<15y : unspecified = female\n",
      "15y-25y : unspecified > female\n",
      "25y-35y : female > unspecified\n",
      "35y-45y : female > unspecified\n",
      "45-75y : unspecified = female\n",
      "\n",
      "age vs product target gender country=France\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male', 'unisex']\n",
      "<15y : unspecified > female\n",
      "15y-25y : female > unspecified\n",
      "25y-35y : unspecified > female\n",
      "35y-45y : female = unspecified\n",
      "45-75y : female = unspecified\n",
      "\n",
      "age vs product target gender country=Germany\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "<15y : unspecified > female > unisex > male\n",
      "15y-25y : unspecified > unisex > female > male\n",
      "25y-35y : male > unisex = female > unspecified\n",
      "35y-45y : female = unisex > unspecified = male\n",
      "45-75y : female > unspecified > unisex = male\n",
      "\n",
      "age vs product target gender country=Italy\n",
      "product target gender categories : ['male', 'unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "<15y : unspecified > female > male\n",
      "15y-25y : unspecified > female > male\n",
      "25y-35y : male > female > unspecified\n",
      "35y-45y : female > unspecified > male\n",
      "45-75y : female > male = unspecified\n",
      "\n",
      "age vs product target gender country=Spain\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male']\n",
      "<15y : unspecified > female\n",
      "15y-25y : unspecified > female\n",
      "25y-35y : female > unspecified\n",
      "35y-45y : female > unspecified\n",
      "45-75y : female = unspecified\n",
      "\n",
      "age vs product target gender country=UK\n",
      "product target gender categories : ['female', 'unspecified', 'male', 'unisex']\n",
      "removed categories : []\n",
      "<15y : unisex > unspecified > female > male\n",
      "15y-25y : unspecified > female > unisex > male\n",
      "25y-35y : male > female > unspecified > unisex\n",
      "35y-45y : male > female > unspecified > unisex\n",
      "45-75y : female > unspecified > male > unisex\n",
      "\n",
      "age vs country product target gender=female\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'Indonesia', 'India']\n",
      "<15y : Germany = USA > UK > France > Spain = Italy\n",
      "15y-25y : France > USA = Germany > UK > Italy > Spain\n",
      "25y-35y : Spain > Italy > UK = USA > Germany = France\n",
      "35y-45y : Spain = Italy > USA = Germany > France = UK\n",
      "45-75y : UK > Germany = Italy > France = Spain > USA\n",
      "\n",
      "age vs country product target gender=male\n",
      "country categories : ['Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'France', 'Spain', 'India', 'Indonesia']\n",
      "<15y : UK > Italy = Germany\n",
      "15y-25y : UK > Italy = Germany\n",
      "25y-35y : Germany = Italy > UK\n",
      "35y-45y : UK > Germany = Italy\n",
      "45-75y : Italy = Germany = UK\n",
      "\n",
      "age vs country product target gender=unspecified\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Indonesia']\n",
      "<15y : Germany > Italy > Spain > USA > France > UK > Brazil > Mexico > India\n",
      "15y-25y : Brazil = USA > UK > Germany > Italy > France > Spain > Mexico > India\n",
      "25y-35y : India > Mexico > Spain > France > Italy > UK > Germany > Brazil > USA\n",
      "35y-45y : Mexico > India > France > Germany > Spain > Italy = UK = USA > Brazil\n",
      "45-75y : Germany > France = Spain = UK > Mexico = Italy > USA > India > Brazil\n",
      "\n",
      "age vs country product target gender=unisex\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'USA', 'France', 'Italy']\n",
      "<15y : UK > Germany\n",
      "15y-25y : Germany > UK\n",
      "25y-35y : Germany > UK\n",
      "35y-45y : Germany > UK\n",
      "45-75y : Germany > UK\n",
      "\n",
      "skin tone vs product target gender country=USA\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex', 'male']\n",
      "light : female > unspecified\n",
      "medium : female = unspecified\n",
      "dark : unspecified > female\n",
      "\n",
      "skin tone vs product target gender country=France\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male', 'unisex']\n",
      "light : female > unspecified\n",
      "medium : unspecified > female\n",
      "dark : female > unspecified\n",
      "\n",
      "skin tone vs product target gender country=Germany\n",
      "product target gender categories : ['female', 'male', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "light : unisex = unspecified > female > male\n",
      "medium : female > unisex = unspecified = male\n",
      "dark : male > female > unspecified > unisex\n",
      "\n",
      "skin tone vs product target gender country=Italy\n",
      "product target gender categories : ['male', 'unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "light : female > unspecified = male\n",
      "medium : unspecified = male > female\n",
      "dark : male > female = unspecified\n",
      "\n",
      "skin tone vs product target gender country=Spain\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male']\n",
      "light : female = unspecified\n",
      "medium : unspecified = female\n",
      "dark : unspecified = female\n",
      "\n",
      "skin tone vs product target gender country=UK\n",
      "product target gender categories : ['female', 'unspecified', 'male', 'unisex']\n",
      "removed categories : []\n",
      "light : unisex > male > unspecified = female\n",
      "medium : unspecified > male = female > unisex\n",
      "dark : female > unspecified > male > unisex\n",
      "\n",
      "skin tone vs country product target gender=female\n",
      "country categories : ['USA', 'France', 'Germany', 'Italy', 'Spain', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'Canada', 'Indonesia', 'India']\n",
      "light : Spain = Italy > France > Germany > USA > UK\n",
      "medium : USA > UK = Germany > Spain = France > Italy\n",
      "dark : UK > France = USA = Germany = Italy > Spain\n",
      "\n",
      "skin tone vs country product target gender=male\n",
      "country categories : ['Germany', 'Italy', 'UK']\n",
      "removed categories : ['Mexico', 'Brazil', 'USA', 'France', 'Spain', 'India', 'Indonesia']\n",
      "light : Italy = UK > Germany\n",
      "medium : UK > Germany = Italy\n",
      "dark : Germany > Italy > UK\n",
      "\n",
      "skin tone vs country product target gender=unspecified\n",
      "country categories : ['Mexico', 'Brazil', 'USA', 'France', 'Germany', 'Italy', 'Spain', 'UK', 'India']\n",
      "removed categories : ['Indonesia']\n",
      "light : India > Mexico > Spain > Germany > Italy > France > Brazil > UK > USA\n",
      "medium : UK > Brazil = USA > France > Germany > Italy > Spain > Mexico > India\n",
      "dark : USA > UK > Brazil > France = Italy > Germany > Spain > Mexico > India\n",
      "\n",
      "skin tone vs country product target gender=unisex\n",
      "country categories : ['Germany', 'UK']\n",
      "removed categories : ['Brazil', 'USA', 'France', 'Italy']\n",
      "light : UK = Germany\n",
      "medium : Germany = UK\n",
      "dark : Germany = UK\n",
      "\n",
      "gender vs product type brand=ARMANI\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "male : fragrance > makeup\n",
      "female : makeup > fragrance\n",
      "\n",
      "gender vs product type brand=GARNIER\n",
      "product type categories : ['haircare', 'makeup', 'skincare']\n",
      "removed categories : ['fragrance']\n",
      "male : skincare > makeup = haircare\n",
      "female : haircare = makeup > skincare\n",
      "\n",
      "gender vs product type brand=LANCOME\n",
      "product type categories : ['fragrance', 'skincare', 'makeup']\n",
      "removed categories : []\n",
      "male : fragrance > skincare > makeup\n",
      "female : makeup > skincare > fragrance\n",
      "\n",
      "gender vs product type brand=LRP\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "male : makeup = skincare\n",
      "female : skincare = makeup\n",
      "\n",
      "gender vs product type brand=MNY\n",
      "product type categories : ['makeup', 'skincare']\n",
      "removed categories : []\n",
      "male : skincare = makeup\n",
      "female : makeup = skincare\n",
      "\n",
      "gender vs product type brand=OAP\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : ['fragrance']\n",
      "male : skincare > makeup > haircare\n",
      "female : haircare > makeup > skincare\n",
      "\n",
      "gender vs product type brand=VICHY\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "male : makeup = haircare > skincare\n",
      "female : skincare > haircare = makeup\n",
      "\n",
      "gender vs product type brand=YSL\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "male : fragrance = makeup\n",
      "female : makeup = fragrance\n",
      "\n",
      "gender vs product type brand=KIEHLS\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "male : skincare > makeup\n",
      "female : makeup > skincare\n",
      "\n",
      "gender vs brand product type=fragrance\n",
      "brand categories : ['ARMANI', 'LANCOME', 'YSL', 'PRADA']\n",
      "removed categories : ['GARNIER', 'OAP']\n",
      "male : PRADA > ARMANI > YSL > LANCOME\n",
      "female : LANCOME > YSL > ARMANI > PRADA\n",
      "\n",
      "gender vs brand product type=haircare\n",
      "brand categories : ['GARNIER', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'REDKEN']\n",
      "removed categories : []\n",
      "male : VICHY > KERASTASE > GARNIER > LOP > OAP = REDKEN\n",
      "female : REDKEN = OAP > LOP > GARNIER > KERASTASE > VICHY\n",
      "\n",
      "gender vs brand product type=makeup\n",
      "brand categories : ['GARNIER', 'MNY', 'OAP', 'NYX', 'YSL', 'ARMANI', 'BIOTHERM', 'KIEHLS', 'LANCOME', 'LRP', 'VICHY']\n",
      "removed categories : ['KERASTASE', 'LOP', 'CERAVE']\n",
      "male : VICHY > LRP > YSL > NYX > OAP > BIOTHERM > GARNIER > KIEHLS > ARMANI > MNY > LANCOME\n",
      "female : LANCOME > MNY > ARMANI > KIEHLS > GARNIER > BIOTHERM > OAP > NYX > YSL > LRP > VICHY\n",
      "\n",
      "gender vs brand product type=skincare\n",
      "brand categories : ['GARNIER', 'LRP', 'OAP', 'VICHY', 'CERAVE', 'KIEHLS', 'LANCOME', 'MNY']\n",
      "removed categories : ['YSL', 'BIOTHERM', 'ARMANI']\n",
      "male : OAP > LRP > KIEHLS > CERAVE = VICHY > GARNIER > LANCOME > MNY\n",
      "female : MNY > LANCOME > GARNIER > VICHY = CERAVE > KIEHLS > LRP > OAP\n",
      "\n",
      "age vs product type brand=ARMANI\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "<15y : fragrance = makeup\n",
      "15y-25y : makeup > fragrance\n",
      "25y-35y : fragrance > makeup\n",
      "35y-45y : fragrance > makeup\n",
      "45-75y : fragrance = makeup\n",
      "\n",
      "age vs product type brand=GARNIER\n",
      "product type categories : ['haircare', 'makeup', 'skincare']\n",
      "removed categories : ['fragrance']\n",
      "<15y : makeup > haircare > skincare\n",
      "15y-25y : skincare > makeup > haircare\n",
      "25y-35y : haircare = skincare > makeup\n",
      "35y-45y : haircare > skincare = makeup\n",
      "45-75y : skincare > makeup = haircare\n",
      "\n",
      "age vs product type brand=LANCOME\n",
      "product type categories : ['fragrance', 'skincare', 'makeup']\n",
      "removed categories : []\n",
      "<15y : makeup > skincare = fragrance\n",
      "15y-25y : makeup > skincare = fragrance\n",
      "25y-35y : fragrance = skincare > makeup\n",
      "35y-45y : fragrance > skincare > makeup\n",
      "45-75y : fragrance > skincare = makeup\n",
      "\n",
      "age vs product type brand=LRP\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "<15y : skincare = makeup\n",
      "15y-25y : skincare > makeup\n",
      "25y-35y : makeup > skincare\n",
      "35y-45y : makeup = skincare\n",
      "45-75y : skincare > makeup\n",
      "\n",
      "age vs product type brand=MNY\n",
      "product type categories : ['makeup', 'skincare']\n",
      "removed categories : []\n",
      "<15y : makeup > skincare\n",
      "15y-25y : makeup > skincare\n",
      "25y-35y : skincare > makeup\n",
      "35y-45y : makeup = skincare\n",
      "45-75y : skincare = makeup\n",
      "\n",
      "age vs product type brand=OAP\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : ['fragrance']\n",
      "<15y : makeup = skincare = haircare\n",
      "15y-25y : makeup = haircare > skincare\n",
      "25y-35y : skincare > haircare > makeup\n",
      "35y-45y : skincare > makeup > haircare\n",
      "45-75y : makeup > skincare > haircare\n",
      "\n",
      "age vs product type brand=VICHY\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "<15y : skincare > makeup > haircare\n",
      "15y-25y : skincare > makeup > haircare\n",
      "25y-35y : haircare > makeup > skincare\n",
      "35y-45y : skincare > haircare = makeup\n",
      "45-75y : skincare > makeup = haircare\n",
      "\n",
      "age vs product type brand=YSL\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "<15y : makeup > fragrance\n",
      "15y-25y : makeup > fragrance\n",
      "25y-35y : fragrance > makeup\n",
      "35y-45y : makeup = fragrance\n",
      "45-75y : makeup > fragrance\n",
      "\n",
      "age vs product type brand=KIEHLS\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "<15y : skincare = makeup\n",
      "15y-25y : skincare > makeup\n",
      "25y-35y : skincare > makeup\n",
      "35y-45y : makeup > skincare\n",
      "45-75y : makeup > skincare\n",
      "\n",
      "age vs brand product type=fragrance\n",
      "brand categories : ['ARMANI', 'LANCOME', 'YSL', 'PRADA']\n",
      "removed categories : ['GARNIER', 'OAP']\n",
      "<15y : PRADA > ARMANI > YSL > LANCOME\n",
      "15y-25y : YSL > ARMANI > PRADA = LANCOME\n",
      "25y-35y : PRADA = LANCOME > ARMANI > YSL\n",
      "35y-45y : LANCOME > ARMANI > PRADA = YSL\n",
      "45-75y : ARMANI = LANCOME > YSL > PRADA\n",
      "\n",
      "age vs brand product type=haircare\n",
      "brand categories : ['GARNIER', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'REDKEN']\n",
      "removed categories : []\n",
      "<15y : GARNIER > LOP > KERASTASE > REDKEN = OAP > VICHY\n",
      "15y-25y : LOP = REDKEN > KERASTASE > GARNIER > OAP > VICHY\n",
      "25y-35y : VICHY > OAP > GARNIER > KERASTASE > LOP > REDKEN\n",
      "35y-45y : REDKEN > KERASTASE = OAP > VICHY = GARNIER > LOP\n",
      "45-75y : KERASTASE > OAP = VICHY = GARNIER > LOP > REDKEN\n",
      "\n",
      "age vs brand product type=makeup\n",
      "brand categories : ['GARNIER', 'MNY', 'OAP', 'NYX', 'YSL', 'ARMANI', 'BIOTHERM', 'KIEHLS', 'LANCOME', 'LRP', 'VICHY']\n",
      "removed categories : ['KERASTASE', 'LOP', 'CERAVE']\n",
      "<15y : GARNIER > NYX > MNY > YSL > LRP > KIEHLS > OAP > LANCOME > ARMANI > VICHY > BIOTHERM\n",
      "15y-25y : MNY > YSL > LANCOME > NYX > ARMANI > GARNIER > BIOTHERM > LRP > OAP > VICHY > KIEHLS\n",
      "25y-35y : VICHY > BIOTHERM > OAP > LRP > ARMANI > KIEHLS > LANCOME > GARNIER > NYX > YSL > MNY\n",
      "35y-45y : KIEHLS > OAP > BIOTHERM > LRP > YSL > NYX > MNY > VICHY > GARNIER > LANCOME > ARMANI\n",
      "45-75y : KIEHLS > OAP > YSL > ARMANI > BIOTHERM > VICHY > LRP > MNY > NYX > LANCOME > GARNIER\n",
      "\n",
      "age vs brand product type=skincare\n",
      "brand categories : ['GARNIER', 'LRP', 'OAP', 'VICHY', 'CERAVE', 'KIEHLS', 'LANCOME', 'MNY']\n",
      "removed categories : ['YSL', 'BIOTHERM', 'ARMANI']\n",
      "<15y : CERAVE > LRP > GARNIER > KIEHLS > MNY = VICHY > OAP > LANCOME\n",
      "15y-25y : LRP > MNY = GARNIER > KIEHLS > VICHY > CERAVE > LANCOME > OAP\n",
      "25y-35y : OAP > LANCOME > CERAVE > KIEHLS > GARNIER = MNY = VICHY > LRP\n",
      "35y-45y : OAP > VICHY > KIEHLS > LANCOME > CERAVE = LRP > MNY > GARNIER\n",
      "45-75y : VICHY > KIEHLS > OAP > MNY = LRP > GARNIER > LANCOME > CERAVE\n",
      "\n",
      "skin tone vs product type brand=ARMANI\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "light : fragrance > makeup\n",
      "medium : makeup > fragrance\n",
      "dark : makeup > fragrance\n",
      "\n",
      "skin tone vs product type brand=GARNIER\n",
      "product type categories : ['haircare', 'makeup', 'skincare']\n",
      "removed categories : ['fragrance']\n",
      "light : makeup > haircare > skincare\n",
      "medium : haircare > skincare > makeup\n",
      "dark : skincare > makeup > haircare\n",
      "\n",
      "skin tone vs product type brand=LANCOME\n",
      "product type categories : ['fragrance', 'skincare', 'makeup']\n",
      "removed categories : []\n",
      "light : fragrance > skincare > makeup\n",
      "medium : makeup > skincare > fragrance\n",
      "dark : makeup > skincare = fragrance\n",
      "\n",
      "skin tone vs product type brand=LRP\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "light : makeup > skincare\n",
      "medium : skincare > makeup\n",
      "dark : skincare > makeup\n",
      "\n",
      "skin tone vs product type brand=MNY\n",
      "product type categories : ['makeup', 'skincare']\n",
      "removed categories : []\n",
      "light : skincare = makeup\n",
      "medium : makeup = skincare\n",
      "dark : skincare = makeup\n",
      "\n",
      "skin tone vs product type brand=OAP\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : ['fragrance']\n",
      "light : skincare > haircare > makeup\n",
      "medium : haircare > skincare = makeup\n",
      "dark : makeup > skincare = haircare\n",
      "\n",
      "skin tone vs product type brand=VICHY\n",
      "product type categories : ['skincare', 'haircare', 'makeup']\n",
      "removed categories : []\n",
      "light : makeup = haircare > skincare\n",
      "medium : skincare > haircare = makeup\n",
      "dark : skincare > makeup = haircare\n",
      "\n",
      "skin tone vs product type brand=YSL\n",
      "product type categories : ['fragrance', 'makeup']\n",
      "removed categories : ['skincare']\n",
      "light : fragrance > makeup\n",
      "medium : makeup > fragrance\n",
      "dark : makeup = fragrance\n",
      "\n",
      "skin tone vs product type brand=KIEHLS\n",
      "product type categories : ['skincare', 'makeup']\n",
      "removed categories : []\n",
      "light : makeup > skincare\n",
      "medium : skincare > makeup\n",
      "dark : skincare > makeup\n",
      "\n",
      "skin tone vs brand product type=fragrance\n",
      "brand categories : ['ARMANI', 'LANCOME', 'YSL', 'PRADA']\n",
      "removed categories : ['GARNIER', 'OAP']\n",
      "light : LANCOME > PRADA > ARMANI = YSL\n",
      "medium : ARMANI > YSL = PRADA > LANCOME\n",
      "dark : YSL > ARMANI > PRADA > LANCOME\n",
      "\n",
      "skin tone vs brand product type=haircare\n",
      "brand categories : ['GARNIER', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'REDKEN']\n",
      "removed categories : []\n",
      "light : VICHY > LOP > KERASTASE > REDKEN = GARNIER > OAP\n",
      "medium : OAP = REDKEN = GARNIER > LOP > KERASTASE > VICHY\n",
      "dark : KERASTASE > OAP > GARNIER > LOP > REDKEN > VICHY\n",
      "\n",
      "skin tone vs brand product type=makeup\n",
      "brand categories : ['GARNIER', 'MNY', 'OAP', 'NYX', 'YSL', 'ARMANI', 'BIOTHERM', 'KIEHLS', 'LANCOME', 'LRP', 'VICHY']\n",
      "removed categories : ['KERASTASE', 'LOP', 'CERAVE']\n",
      "light : VICHY > BIOTHERM > LRP > GARNIER > KIEHLS > MNY > OAP > YSL > NYX > LANCOME > ARMANI\n",
      "medium : LANCOME > ARMANI > NYX > MNY > KIEHLS > OAP > YSL > GARNIER > LRP > BIOTHERM > VICHY\n",
      "dark : ARMANI > YSL > NYX > OAP > LANCOME > LRP > BIOTHERM > MNY > GARNIER > KIEHLS > VICHY\n",
      "\n",
      "skin tone vs brand product type=skincare\n",
      "brand categories : ['GARNIER', 'LRP', 'OAP', 'VICHY', 'CERAVE', 'KIEHLS', 'LANCOME', 'MNY']\n",
      "removed categories : ['YSL', 'BIOTHERM', 'ARMANI']\n",
      "light : CERAVE > LANCOME > VICHY > OAP > MNY > GARNIER > LRP > KIEHLS\n",
      "medium : KIEHLS > LRP > OAP = MNY > GARNIER = VICHY > LANCOME > CERAVE\n",
      "dark : LRP = GARNIER > KIEHLS > MNY > LANCOME = VICHY > OAP > CERAVE\n",
      "\n",
      "gender vs product target gender brand=ARMANI\n",
      "product target gender categories : ['female', 'male', 'unspecified']\n",
      "removed categories : []\n",
      "male : male > unspecified > female\n",
      "female : female > unspecified > male\n",
      "\n",
      "gender vs product target gender brand=GARNIER\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male', 'unisex']\n",
      "male : female = unspecified\n",
      "female : unspecified = female\n",
      "\n",
      "gender vs product target gender brand=LANCOME\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : []\n",
      "male : female > unspecified\n",
      "female : unspecified > female\n",
      "\n",
      "gender vs product target gender brand=LRP\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "male : unspecified > female\n",
      "female : female > unspecified\n",
      "\n",
      "gender vs product target gender brand=OAP\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "male : male > unspecified = female\n",
      "female : female = unspecified > male\n",
      "\n",
      "gender vs product target gender brand=VICHY\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['unisex']\n",
      "male : unspecified > female\n",
      "female : female > unspecified\n",
      "\n",
      "gender vs product target gender brand=YSL\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male']\n",
      "male : unspecified > female\n",
      "female : female > unspecified\n",
      "\n",
      "gender vs product target gender brand=CERAVE\n",
      "product target gender categories : ['unisex', 'female']\n",
      "removed categories : ['unspecified']\n",
      "male : unisex > female\n",
      "female : female > unisex\n",
      "\n",
      "gender vs product target gender brand=KIEHLS\n",
      "product target gender categories : ['female', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "male : female = unisex = unspecified\n",
      "female : unspecified = unisex = female\n",
      "\n",
      "gender vs brand product target gender=female\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KIEHLS', 'LRP']\n",
      "removed categories : ['PRADA', 'BIOTHERM']\n",
      "male : LRP > KIEHLS > VICHY > GARNIER > YSL = LANCOME > CERAVE > ARMANI > OAP\n",
      "female : OAP > ARMANI > CERAVE > LANCOME = YSL > GARNIER > VICHY > KIEHLS > LRP\n",
      "\n",
      "gender vs brand product target gender=male\n",
      "brand categories : ['ARMANI', 'OAP', 'PRADA']\n",
      "removed categories : ['GARNIER', 'YSL', 'BIOTHERM']\n",
      "male : PRADA = OAP > ARMANI\n",
      "female : ARMANI > OAP = PRADA\n",
      "\n",
      "gender vs brand product target gender=unspecified\n",
      "brand categories : ['GARNIER', 'LRP', 'MNY', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'ARMANI', 'NYX', 'REDKEN', 'YSL', 'BIOTHERM', 'KIEHLS', 'LANCOME']\n",
      "removed categories : ['CERAVE', 'PRADA']\n",
      "male : VICHY > LRP > YSL > NYX > KERASTASE > ARMANI > BIOTHERM > KIEHLS > GARNIER > LOP > REDKEN > OAP > MNY > LANCOME\n",
      "female : LANCOME > MNY > OAP > REDKEN > LOP > GARNIER > KIEHLS > BIOTHERM > ARMANI > KERASTASE > NYX > YSL > LRP > VICHY\n",
      "\n",
      "gender vs brand product target gender=unisex\n",
      "brand categories : ['CERAVE', 'KIEHLS']\n",
      "removed categories : ['GARNIER', 'LRP', 'VICHY']\n",
      "male : CERAVE = KIEHLS\n",
      "female : KIEHLS = CERAVE\n",
      "\n",
      "age vs product target gender brand=ARMANI\n",
      "product target gender categories : ['female', 'male', 'unspecified']\n",
      "removed categories : []\n",
      "<15y : female = unspecified > male\n",
      "15y-25y : unspecified > female > male\n",
      "25y-35y : male > female > unspecified\n",
      "35y-45y : male = female > unspecified\n",
      "45-75y : unspecified = female = male\n",
      "\n",
      "age vs product target gender brand=GARNIER\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male', 'unisex']\n",
      "<15y : unspecified > female\n",
      "15y-25y : female > unspecified\n",
      "25y-35y : female = unspecified\n",
      "35y-45y : unspecified > female\n",
      "45-75y : female > unspecified\n",
      "\n",
      "age vs product target gender brand=LANCOME\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : []\n",
      "<15y : unspecified > female\n",
      "15y-25y : unspecified > female\n",
      "25y-35y : female > unspecified\n",
      "35y-45y : female > unspecified\n",
      "45-75y : female = unspecified\n",
      "\n",
      "age vs product target gender brand=LRP\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "<15y : female > unspecified\n",
      "15y-25y : female > unspecified\n",
      "25y-35y : unspecified > female\n",
      "35y-45y : female > unspecified\n",
      "45-75y : female > unspecified\n",
      "\n",
      "age vs product target gender brand=OAP\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "<15y : male = unspecified > female\n",
      "15y-25y : unspecified > male > female\n",
      "25y-35y : female > male > unspecified\n",
      "35y-45y : male > female > unspecified\n",
      "45-75y : female > unspecified > male\n",
      "\n",
      "age vs product target gender brand=VICHY\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['unisex']\n",
      "<15y : female > unspecified\n",
      "15y-25y : female > unspecified\n",
      "25y-35y : unspecified > female\n",
      "35y-45y : female > unspecified\n",
      "45-75y : female > unspecified\n",
      "\n",
      "age vs product target gender brand=YSL\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male']\n",
      "<15y : unspecified > female\n",
      "15y-25y : unspecified = female\n",
      "25y-35y : female > unspecified\n",
      "35y-45y : unspecified > female\n",
      "45-75y : unspecified = female\n",
      "\n",
      "age vs product target gender brand=CERAVE\n",
      "product target gender categories : ['unisex', 'female']\n",
      "removed categories : ['unspecified']\n",
      "<15y : unisex > female\n",
      "15y-25y : unisex = female\n",
      "25y-35y : female > unisex\n",
      "35y-45y : female = unisex\n",
      "45-75y : unisex = female\n",
      "\n",
      "age vs product target gender brand=KIEHLS\n",
      "product target gender categories : ['female', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "<15y : female = unspecified = unisex\n",
      "15y-25y : unisex > female > unspecified\n",
      "25y-35y : female > unisex = unspecified\n",
      "35y-45y : unspecified > unisex > female\n",
      "45-75y : unspecified > female > unisex\n",
      "\n",
      "age vs brand product target gender=female\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KIEHLS', 'LRP']\n",
      "removed categories : ['PRADA', 'BIOTHERM']\n",
      "<15y : LRP > CERAVE > GARNIER = KIEHLS > ARMANI > VICHY > YSL > LANCOME = OAP\n",
      "15y-25y : LRP = YSL > GARNIER > ARMANI > KIEHLS > VICHY > CERAVE > LANCOME > OAP\n",
      "25y-35y : OAP > LANCOME > CERAVE > ARMANI = KIEHLS > YSL = GARNIER = VICHY > LRP\n",
      "35y-45y : OAP > VICHY > LANCOME > KIEHLS > ARMANI > LRP = CERAVE > GARNIER > YSL\n",
      "45-75y : VICHY > KIEHLS > OAP > LRP > ARMANI > GARNIER > LANCOME > YSL > CERAVE\n",
      "\n",
      "age vs brand product target gender=male\n",
      "brand categories : ['ARMANI', 'OAP', 'PRADA']\n",
      "removed categories : ['GARNIER', 'YSL', 'BIOTHERM']\n",
      "<15y : OAP = PRADA = ARMANI\n",
      "15y-25y : OAP > ARMANI > PRADA\n",
      "25y-35y : PRADA > ARMANI > OAP\n",
      "35y-45y : OAP > ARMANI = PRADA\n",
      "45-75y : ARMANI = OAP > PRADA\n",
      "\n",
      "age vs brand product target gender=unspecified\n",
      "brand categories : ['GARNIER', 'LRP', 'MNY', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'ARMANI', 'NYX', 'REDKEN', 'YSL', 'BIOTHERM', 'KIEHLS', 'LANCOME']\n",
      "removed categories : ['CERAVE', 'PRADA']\n",
      "<15y : GARNIER > NYX > LOP > MNY > YSL > KERASTASE > KIEHLS > ARMANI > REDKEN > LRP > LANCOME > OAP > BIOTHERM > VICHY\n",
      "15y-25y : LOP > REDKEN > MNY > LANCOME > NYX > YSL > ARMANI > KERASTASE > LRP > GARNIER > BIOTHERM > OAP > VICHY > KIEHLS\n",
      "25y-35y : VICHY > OAP > BIOTHERM > LRP > KIEHLS > GARNIER > ARMANI > KERASTASE > LANCOME > YSL > NYX > MNY > LOP > REDKEN\n",
      "35y-45y : KIEHLS > REDKEN > OAP > BIOTHERM > KERASTASE > YSL > NYX > VICHY > MNY > GARNIER > LOP > ARMANI > LANCOME > LRP\n",
      "45-75y : KIEHLS > ARMANI > OAP > YSL > KERASTASE > BIOTHERM > VICHY > MNY > NYX > LANCOME > GARNIER > LOP > LRP > REDKEN\n",
      "\n",
      "age vs brand product target gender=unisex\n",
      "brand categories : ['CERAVE', 'KIEHLS']\n",
      "removed categories : ['GARNIER', 'LRP', 'VICHY']\n",
      "<15y : CERAVE > KIEHLS\n",
      "15y-25y : KIEHLS > CERAVE\n",
      "25y-35y : CERAVE = KIEHLS\n",
      "35y-45y : KIEHLS > CERAVE\n",
      "45-75y : KIEHLS > CERAVE\n",
      "\n",
      "skin tone vs product target gender brand=ARMANI\n",
      "product target gender categories : ['female', 'male', 'unspecified']\n",
      "removed categories : []\n",
      "light : female = male > unspecified\n",
      "medium : unspecified > female = male\n",
      "dark : unspecified = male > female\n",
      "\n",
      "skin tone vs product target gender brand=GARNIER\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['male', 'unisex']\n",
      "light : unspecified > female\n",
      "medium : unspecified = female\n",
      "dark : female > unspecified\n",
      "\n",
      "skin tone vs product target gender brand=LANCOME\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : []\n",
      "light : female > unspecified\n",
      "medium : unspecified > female\n",
      "dark : unspecified > female\n",
      "\n",
      "skin tone vs product target gender brand=LRP\n",
      "product target gender categories : ['unspecified', 'female']\n",
      "removed categories : ['unisex']\n",
      "light : unspecified > female\n",
      "medium : female > unspecified\n",
      "dark : unspecified = female\n",
      "\n",
      "skin tone vs product target gender brand=OAP\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "light : female > male = unspecified\n",
      "medium : male > unspecified > female\n",
      "dark : unspecified > female = male\n",
      "\n",
      "skin tone vs product target gender brand=VICHY\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['unisex']\n",
      "light : unspecified > female\n",
      "medium : female > unspecified\n",
      "dark : female > unspecified\n",
      "\n",
      "skin tone vs product target gender brand=YSL\n",
      "product target gender categories : ['female', 'unspecified']\n",
      "removed categories : ['male']\n",
      "light : female > unspecified\n",
      "medium : unspecified > female\n",
      "dark : unspecified > female\n",
      "\n",
      "skin tone vs product target gender brand=CERAVE\n",
      "product target gender categories : ['unisex', 'female']\n",
      "removed categories : ['unspecified']\n",
      "light : female = unisex\n",
      "medium : unisex = female\n",
      "dark : unisex = female\n",
      "\n",
      "skin tone vs product target gender brand=KIEHLS\n",
      "product target gender categories : ['female', 'unspecified', 'unisex']\n",
      "removed categories : []\n",
      "light : unspecified > unisex = female\n",
      "medium : female = unisex > unspecified\n",
      "dark : female > unisex > unspecified\n",
      "\n",
      "skin tone vs brand product target gender=female\n",
      "brand categories : ['ARMANI', 'GARNIER', 'LANCOME', 'OAP', 'VICHY', 'YSL', 'CERAVE', 'KIEHLS', 'LRP']\n",
      "removed categories : ['PRADA', 'BIOTHERM']\n",
      "light : CERAVE > LANCOME > OAP > VICHY > YSL = ARMANI > GARNIER > LRP > KIEHLS\n",
      "medium : KIEHLS > LRP > ARMANI = VICHY > GARNIER > OAP > YSL > LANCOME > CERAVE\n",
      "dark : YSL > GARNIER > LRP > KIEHLS > ARMANI > OAP > VICHY = LANCOME > CERAVE\n",
      "\n",
      "skin tone vs brand product target gender=male\n",
      "brand categories : ['ARMANI', 'OAP', 'PRADA']\n",
      "removed categories : ['GARNIER', 'YSL', 'BIOTHERM']\n",
      "light : PRADA > OAP = ARMANI\n",
      "medium : OAP > ARMANI > PRADA\n",
      "dark : ARMANI > PRADA > OAP\n",
      "\n",
      "skin tone vs brand product target gender=unspecified\n",
      "brand categories : ['GARNIER', 'LRP', 'MNY', 'OAP', 'VICHY', 'KERASTASE', 'LOP', 'ARMANI', 'NYX', 'REDKEN', 'YSL', 'BIOTHERM', 'KIEHLS', 'LANCOME']\n",
      "removed categories : ['CERAVE', 'PRADA']\n",
      "light : VICHY > LOP > BIOTHERM > KERASTASE > GARNIER > REDKEN > LRP > KIEHLS > OAP > MNY > YSL > NYX > LANCOME > ARMANI\n",
      "medium : LANCOME > ARMANI > NYX > KIEHLS > MNY > REDKEN > OAP > GARNIER > YSL > LOP > LRP > KERASTASE > BIOTHERM > VICHY\n",
      "dark : YSL > ARMANI > NYX > LRP > LANCOME > OAP > BIOTHERM > MNY > KERASTASE > GARNIER > KIEHLS > LOP > REDKEN > VICHY\n",
      "\n",
      "skin tone vs brand product target gender=unisex\n",
      "brand categories : ['CERAVE', 'KIEHLS']\n",
      "removed categories : ['GARNIER', 'LRP', 'VICHY']\n",
      "light : CERAVE > KIEHLS\n",
      "medium : KIEHLS > CERAVE\n",
      "dark : KIEHLS > CERAVE\n",
      "\n",
      "gender vs product target gender product type=fragrance\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "male : male > unspecified > female\n",
      "female : female > unspecified > male\n",
      "\n",
      "gender vs product target gender product type=haircare\n",
      "product target gender categories : ['male', 'unspecified']\n",
      "removed categories : []\n",
      "male : male > unspecified\n",
      "female : unspecified > male\n",
      "\n",
      "gender vs product target gender product type=skincare\n",
      "product target gender categories : ['female', 'unspecified', 'unisex', 'male']\n",
      "removed categories : []\n",
      "male : male > unspecified = unisex > female\n",
      "female : female > unisex = unspecified > male\n",
      "\n",
      "gender vs product type product target gender=female\n",
      "product type categories : ['fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "male : skincare > fragrance\n",
      "female : fragrance > skincare\n",
      "\n",
      "gender vs product type product target gender=male\n",
      "product type categories : ['haircare', 'fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "male : haircare = skincare > fragrance\n",
      "female : fragrance > skincare = haircare\n",
      "\n",
      "gender vs product type product target gender=unspecified\n",
      "product type categories : ['makeup', 'haircare', 'fragrance', 'skincare']\n",
      "removed categories : []\n",
      "male : skincare = fragrance > makeup > haircare\n",
      "female : haircare > makeup > fragrance = skincare\n",
      "\n",
      "age vs product target gender product type=fragrance\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "<15y : unspecified > female > male\n",
      "15y-25y : unspecified = female > male\n",
      "25y-35y : male > female > unspecified\n",
      "35y-45y : female = male = unspecified\n",
      "45-75y : unspecified = female > male\n",
      "\n",
      "age vs product target gender product type=haircare\n",
      "product target gender categories : ['male', 'unspecified']\n",
      "removed categories : []\n",
      "<15y : unspecified > male\n",
      "15y-25y : unspecified > male\n",
      "25y-35y : male > unspecified\n",
      "35y-45y : male > unspecified\n",
      "45-75y : male > unspecified\n",
      "\n",
      "age vs product target gender product type=skincare\n",
      "product target gender categories : ['female', 'unspecified', 'unisex', 'male']\n",
      "removed categories : []\n",
      "<15y : unisex > female > unspecified = male\n",
      "15y-25y : unspecified > unisex = female > male\n",
      "25y-35y : male > female = unspecified > unisex\n",
      "35y-45y : male > female > unisex > unspecified\n",
      "45-75y : female > unisex = unspecified > male\n",
      "\n",
      "age vs product type product target gender=female\n",
      "product type categories : ['fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "<15y : skincare > fragrance\n",
      "15y-25y : skincare > fragrance\n",
      "25y-35y : fragrance > skincare\n",
      "35y-45y : skincare = fragrance\n",
      "45-75y : skincare > fragrance\n",
      "\n",
      "age vs product type product target gender=male\n",
      "product type categories : ['haircare', 'fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "<15y : skincare > fragrance = haircare\n",
      "15y-25y : haircare = skincare > fragrance\n",
      "25y-35y : fragrance > skincare = haircare\n",
      "35y-45y : skincare = haircare > fragrance\n",
      "45-75y : haircare > fragrance > skincare\n",
      "\n",
      "age vs product type product target gender=unspecified\n",
      "product type categories : ['makeup', 'haircare', 'fragrance', 'skincare']\n",
      "removed categories : []\n",
      "<15y : makeup = haircare > fragrance > skincare\n",
      "15y-25y : makeup = skincare > haircare > fragrance\n",
      "25y-35y : fragrance = skincare = haircare > makeup\n",
      "35y-45y : makeup = fragrance = haircare > skincare\n",
      "45-75y : fragrance = makeup > haircare > skincare\n",
      "\n",
      "skin tone vs product target gender product type=fragrance\n",
      "product target gender categories : ['female', 'unspecified', 'male']\n",
      "removed categories : []\n",
      "light : female > male > unspecified\n",
      "medium : unspecified > male > female\n",
      "dark : unspecified = male > female\n",
      "\n",
      "skin tone vs product target gender product type=haircare\n",
      "product target gender categories : ['male', 'unspecified']\n",
      "removed categories : []\n",
      "light : unspecified > male\n",
      "medium : male > unspecified\n",
      "dark : unspecified = male\n",
      "\n",
      "skin tone vs product target gender product type=skincare\n",
      "product target gender categories : ['female', 'unspecified', 'unisex', 'male']\n",
      "removed categories : []\n",
      "light : unisex > male > unspecified = female\n",
      "medium : male > female > unisex = unspecified\n",
      "dark : unspecified > female > unisex > male\n",
      "\n",
      "skin tone vs product type product target gender=female\n",
      "product type categories : ['fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "light : fragrance > skincare\n",
      "medium : skincare > fragrance\n",
      "dark : skincare = fragrance\n",
      "\n",
      "skin tone vs product type product target gender=male\n",
      "product type categories : ['haircare', 'fragrance', 'skincare']\n",
      "removed categories : ['makeup']\n",
      "light : skincare = haircare = fragrance\n",
      "medium : haircare = skincare > fragrance\n",
      "dark : fragrance > haircare = skincare\n",
      "\n",
      "skin tone vs product type product target gender=unspecified\n",
      "product type categories : ['makeup', 'haircare', 'fragrance', 'skincare']\n",
      "removed categories : []\n",
      "light : haircare > skincare > makeup > fragrance\n",
      "medium : fragrance > makeup > haircare > skincare\n",
      "dark : fragrance > skincare > makeup > haircare\n",
      "\n"
     ]
    }
   ],
   "source": [
    "metadata_vars = [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "rp_vars = [\"gender\", \"age\", \"skin tone\"]\n",
    "\n",
    "for i in range(len(metadata_vars)):\n",
    "    for j in range(i + 1, len(metadata_vars)):\n",
    "        for rp in rp_vars:\n",
    "            metadata2_x_rp(data_df, rp_data2, metadata_vars[i], metadata_vars[j], rp)"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Gender x Metadata x Metadata HeatMap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "def gender_heatmap(df, rp_data, metadata1, metadata2, alpha=0.05):\n",
    "    seaborn.set_style(\"dark\")\n",
    "    assert metadata1 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    assert metadata2 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    metadata1_cats_dist = collections.Counter(df[metadata1])\n",
    "    metadata2_cats_dist = collections.Counter(df[metadata2])\n",
    "    metadata1_cats = np.array(sorted(metadata1_cats_dist.keys(), key=lambda cat: metadata1_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    metadata2_cats = np.array(sorted(metadata2_cats_dist.keys(), key=lambda cat: metadata2_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    count_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=int)\n",
    "    for (metadata1_cat, metadata2_cat), gdf in df.groupby([metadata1, metadata2]):\n",
    "        i = (metadata1_cats == metadata1_cat).nonzero()[0].item()\n",
    "        j = (metadata2_cats == metadata2_cat).nonzero()[0].item()\n",
    "        count_arr[i, j] = len(gdf)\n",
    "    row_index = (count_arr >= 30).any(axis=1)\n",
    "    metadata1_cats, count_arr = metadata1_cats[row_index], count_arr[row_index]\n",
    "    column_index = (count_arr >= 30).any(axis=0)\n",
    "    metadata2_cats, count_arr = metadata2_cats[column_index], count_arr[:, column_index]\n",
    "    value_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=float)\n",
    "    for i, metadata1_cat in enumerate(metadata1_cats):\n",
    "        for j, metadata2_cat in enumerate(metadata2_cats):\n",
    "            ind = (df[metadata1] == metadata1_cat) & (df[metadata2] == metadata2_cat)\n",
    "            if count_arr[i, j] < 30:\n",
    "                value_arr[i, j] = np.nan\n",
    "            else:\n",
    "                n_female_faces = rp_data[ind, 1].sum(axis=(1, 2))\n",
    "                n_total_faces = rp_data[ind].sum(axis=(1, 2, 3))\n",
    "                value_arr[i, j] = (n_female_faces/n_total_faces).mean()\n",
    "                res = scipy.stats.ttest_1samp(n_female_faces/n_total_faces, 0.5, alternative=\"two-sided\")\n",
    "                pvalue = res.pvalue\n",
    "                if pvalue >= alpha:\n",
    "                    value_arr[i, j] = 1.5\n",
    "    cmap = matplotlib.colormaps[\"RdBu\"](np.linspace(0.2, 0.9, 100))\n",
    "    cmap = matplotlib.colors.ListedColormap(cmap)\n",
    "    cmap.set_under(\"none\")\n",
    "    cmap2 = matplotlib.colormaps[\"binary\"]\n",
    "    cmap2.set_over(\"none\")\n",
    "    height = len(metadata1_cats) * 1.2\n",
    "    width = len(metadata2_cats)\n",
    "    plt.figure(figsize=(width, height))\n",
    "    ax = seaborn.heatmap(100*value_arr, cmap=cmap2, vmin=100, vmax=200, cbar=False, annot=False)\n",
    "    value_arr[value_arr == 1.5] = np.nan\n",
    "    seaborn.heatmap(100*value_arr, xticklabels=metadata2_cats, yticklabels=metadata1_cats, annot=True, fmt=\".2f\", \n",
    "                    linewidth=1, cmap=cmap, annot_kws={\"fontsize\":\"xx-small\"}, cbar=True, vmin=0, vmax=100, \n",
    "                    linecolor=\"black\", ax=ax)\n",
    "    ax.set_xticklabels(metadata2_cats, fontsize=\"xx-small\")\n",
    "    ax.set_yticklabels(metadata1_cats, fontsize=\"xx-small\")\n",
    "    cbar = ax.collections[1].colorbar\n",
    "    cbar.ax.tick_params(labelsize=\"xx-small\")\n",
    "    ax.set_xlabel(metadata2.upper(), fontsize=\"xx-small\")\n",
    "    ax.set_ylabel(metadata1.upper(), fontsize=\"xx-small\")\n",
    "    ax.set_title(f\"Female Screentime for {metadata1.upper()} x {metadata2.upper()}\", fontsize=\"xx-small\")\n",
    "    filename = f\"plots/rp_x_metadata_x_metadata/gender_x_{metadata1}_x_{metadata2}.png\"\n",
    "    figure = ax.get_figure()\n",
    "    figure.savefig(filename, dpi=300, bbox_inches=\"tight\")\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "gender_heatmap(data_df, rp_data2, \"country\", \"brand\")\n",
    "gender_heatmap(data_df, rp_data2, \"product type\", \"brand\")\n",
    "gender_heatmap(data_df, rp_data2, \"product type\", \"country\")\n",
    "gender_heatmap(data_df, rp_data2, \"product target gender\", \"brand\")\n",
    "gender_heatmap(data_df, rp_data2, \"product target gender\", \"country\")\n",
    "gender_heatmap(data_df, rp_data2, \"product target gender\", \"product type\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Age x Metadata x Metadata HeatMap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def age_heatmap(df, rp_data, metadata1, metadata2, alpha=0.05):\n",
    "    seaborn.set_style(\"dark\")\n",
    "    rp_cats = [\"0-15\", \"15-25\", \"25-35\", \"35-45\", \"45-75\"]\n",
    "    cmap_names = [\"Wistia\", \"Blues\", \"Reds\", \"Greens\", \"Purples\"]\n",
    "    assert metadata1 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    assert metadata2 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    metadata1_cats_dist = collections.Counter(df[metadata1])\n",
    "    metadata2_cats_dist = collections.Counter(df[metadata2])\n",
    "    metadata1_cats = np.array(sorted(metadata1_cats_dist.keys(), key=lambda cat: metadata1_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    metadata2_cats = np.array(sorted(metadata2_cats_dist.keys(), key=lambda cat: metadata2_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    count_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=int)\n",
    "    for (metadata1_cat, metadata2_cat), gdf in df.groupby([metadata1, metadata2]):\n",
    "        i = (metadata1_cats == metadata1_cat).nonzero()[0].item()\n",
    "        j = (metadata2_cats == metadata2_cat).nonzero()[0].item()\n",
    "        count_arr[i, j] = len(gdf)\n",
    "    row_index = (count_arr >= 30).any(axis=1)\n",
    "    metadata1_cats, count_arr = metadata1_cats[row_index], count_arr[row_index]\n",
    "    column_index = (count_arr >= 30).any(axis=0)\n",
    "    metadata2_cats, count_arr = metadata2_cats[column_index], count_arr[:, column_index]\n",
    "\n",
    "    height = len(metadata1_cats) * 1.5\n",
    "    width = len(metadata2_cats) + 15\n",
    "    plt.figure(figsize=(width, height))\n",
    "    ax = plt.gca()\n",
    "\n",
    "    grey_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=float)\n",
    "\n",
    "    for age_cat in range(len(rp_cats) - 1, -1, -1):\n",
    "        rp_cat, cmap_name = rp_cats[age_cat], cmap_names[age_cat]\n",
    "        value_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=float)\n",
    "        for i, metadata1_cat in enumerate(metadata1_cats):\n",
    "            for j, metadata2_cat in enumerate(metadata2_cats):\n",
    "                ind = (df[metadata1] == metadata1_cat) & (df[metadata2] == metadata2_cat)\n",
    "                if count_arr[i, j] < 30:\n",
    "                    value_arr[i, j] = grey_arr[i, j] = np.nan\n",
    "                else:\n",
    "                    n_age_faces = rp_data[ind, :, age_cat].sum(axis=(1, 2))\n",
    "                    n_total_faces = rp_data[ind].sum(axis=(1, 2, 3))\n",
    "                    value_arr[i, j] = (n_age_faces/n_total_faces).mean()\n",
    "                    res = scipy.stats.ttest_1samp(n_age_faces/n_total_faces, 0.5, alternative=\"greater\")\n",
    "                    pvalue = res.pvalue\n",
    "                    if pvalue >= alpha:\n",
    "                        value_arr[i, j] = np.nan\n",
    "                        grey_arr[i, j] = grey_arr[i, j] * grey_arr[i, j]\n",
    "                    else:\n",
    "                        grey_arr[i, j] = np.nan\n",
    "        cmap = matplotlib.colormaps[cmap_name](np.linspace(0.3, 0.9, 100))\n",
    "        cmap = matplotlib.colors.ListedColormap(cmap)\n",
    "        seaborn.heatmap(100*value_arr, xticklabels=metadata2_cats, yticklabels=metadata1_cats, annot=True, fmt=\".2f\",\n",
    "                        linewidth=2, cmap=cmap, annot_kws={\"fontsize\":\"x-small\"}, cbar=True, vmin=50, vmax=100, \n",
    "                        linecolor=\"black\", ax=ax, cbar_kws={\"fraction\": 0.04, \"pad\": 0.025, \"shrink\": 0.75})\n",
    "        cbar = ax.collections[len(rp_cats) - age_cat - 1].colorbar\n",
    "        cbar.ax.tick_params(labelsize=\"xx-small\")\n",
    "        cbar.ax.set_ylabel(f\"{rp_cat}y\", fontsize=\"xx-small\")\n",
    "    \n",
    "    binary_cmap = matplotlib.colormaps[\"binary\"]\n",
    "    grey_arr[grey_arr == 0] = 0.5\n",
    "    seaborn.heatmap(grey_arr, cmap=binary_cmap, cbar=False, annot=False, vmin=0, vmax=1, ax=ax, linewidths=2, \n",
    "                    linecolor=\"black\")\n",
    "\n",
    "    ax.set_xticklabels(metadata2_cats, fontsize=\"x-small\")\n",
    "    ax.set_yticklabels(metadata1_cats, fontsize=\"x-small\")\n",
    "    ax.set_xlabel(metadata2.upper(), fontsize=\"x-small\")\n",
    "    ax.set_ylabel(metadata1.upper(), fontsize=\"x-small\")\n",
    "    ax.set_title(f\"Age Screentime for {metadata1.upper()} x {metadata2.upper()}\", fontsize=\"x-small\")\n",
    "    filename = f\"plots/rp_x_metadata_x_metadata/age_x_{metadata1}_x_{metadata2}.png\"\n",
    "    figure = ax.get_figure()\n",
    "    figure.savefig(filename, dpi=300, bbox_inches=\"tight\")\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "age_heatmap(data_df, rp_data2, \"country\", \"brand\")\n",
    "age_heatmap(data_df, rp_data2, \"product type\", \"brand\")\n",
    "age_heatmap(data_df, rp_data2, \"product type\", \"country\")\n",
    "age_heatmap(data_df, rp_data2, \"product target gender\", \"brand\")\n",
    "age_heatmap(data_df, rp_data2, \"product target gender\", \"country\")\n",
    "age_heatmap(data_df, rp_data2, \"product target gender\", \"product type\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Skin Tone x Metadata x Metadata HeatMap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def skin_tone_heatmap(df, rp_data, metadata1, metadata2, alpha=0.05):\n",
    "    seaborn.set_style(\"dark\")\n",
    "    rp_cats = [\"light\", \"medium\", \"dark\"]\n",
    "    cmap_names = [\"Blues\", \"Wistia\", \"Greens\"]\n",
    "    assert metadata1 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    assert metadata2 in [\"country\", \"brand\", \"product type\", \"product target gender\"]\n",
    "    metadata1_cats_dist = collections.Counter(df[metadata1])\n",
    "    metadata2_cats_dist = collections.Counter(df[metadata2])\n",
    "    metadata1_cats = np.array(sorted(metadata1_cats_dist.keys(), key=lambda cat: metadata1_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    metadata2_cats = np.array(sorted(metadata2_cats_dist.keys(), key=lambda cat: metadata2_cats_dist[cat], \n",
    "                                     reverse=True))\n",
    "    count_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=int)\n",
    "    for (metadata1_cat, metadata2_cat), gdf in df.groupby([metadata1, metadata2]):\n",
    "        i = (metadata1_cats == metadata1_cat).nonzero()[0].item()\n",
    "        j = (metadata2_cats == metadata2_cat).nonzero()[0].item()\n",
    "        count_arr[i, j] = len(gdf)\n",
    "    row_index = (count_arr >= 30).any(axis=1)\n",
    "    metadata1_cats, count_arr = metadata1_cats[row_index], count_arr[row_index]\n",
    "    column_index = (count_arr >= 30).any(axis=0)\n",
    "    metadata2_cats, count_arr = metadata2_cats[column_index], count_arr[:, column_index]\n",
    "\n",
    "    height = len(metadata1_cats) * 1.2\n",
    "    width = len(metadata2_cats) + 10\n",
    "    plt.figure(figsize=(width, height))\n",
    "    ax = plt.gca()\n",
    "\n",
    "    grey_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=float)\n",
    "\n",
    "    for skin_tone_cat in range(len(rp_cats) - 1, -1, -1):\n",
    "        rp_cat, cmap_name = rp_cats[skin_tone_cat], cmap_names[skin_tone_cat]\n",
    "        value_arr = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=float)\n",
    "        for i, metadata1_cat in enumerate(metadata1_cats):\n",
    "            for j, metadata2_cat in enumerate(metadata2_cats):\n",
    "                ind = (df[metadata1] == metadata1_cat) & (df[metadata2] == metadata2_cat)\n",
    "                if count_arr[i, j] < 30:\n",
    "                    value_arr[i, j] = grey_arr[i, j] = np.nan\n",
    "                else:\n",
    "                    n_skin_tone_faces = rp_data[ind, :, :, skin_tone_cat].sum(axis=(1, 2))\n",
    "                    n_total_faces = rp_data[ind].sum(axis=(1, 2, 3))\n",
    "                    value_arr[i, j] = (n_skin_tone_faces/n_total_faces).mean()\n",
    "                    res = scipy.stats.ttest_1samp(n_skin_tone_faces/n_total_faces, 0.5, alternative=\"greater\")\n",
    "                    pvalue = res.pvalue\n",
    "                    if pvalue >= alpha:\n",
    "                        value_arr[i, j] = np.nan\n",
    "                        grey_arr[i, j] = grey_arr[i, j] * grey_arr[i, j]\n",
    "                    else:\n",
    "                        grey_arr[i, j] = np.nan\n",
    "        cmap = matplotlib.colormaps[cmap_name](np.linspace(0.2, 0.9, 100))\n",
    "        cmap = matplotlib.colors.ListedColormap(cmap)\n",
    "        seaborn.heatmap(100*value_arr, xticklabels=metadata2_cats, yticklabels=metadata1_cats, annot=True, fmt=\".2f\",\n",
    "                        linewidth=1, cmap=cmap, annot_kws={\"fontsize\":\"xx-small\"}, cbar=True, vmin=50, vmax=100, \n",
    "                        linecolor=\"black\", ax=ax, cbar_kws={\"fraction\": 0.04, \"pad\": 0.025, \"shrink\": 0.75})\n",
    "        cbar = ax.collections[len(rp_cats) - skin_tone_cat - 1].colorbar\n",
    "        cbar.ax.tick_params(labelsize=\"xx-small\")\n",
    "        cbar.ax.set_ylabel(f\"{rp_cat}\", fontsize=\"xx-small\")\n",
    "\n",
    "    binary_cmap = matplotlib.colormaps[\"binary\"]\n",
    "    grey_arr[grey_arr == 0] = 0.5\n",
    "    seaborn.heatmap(grey_arr, cmap=binary_cmap, cbar=False, annot=False, vmin=0, vmax=1, ax=ax, linewidths=2, \n",
    "                    linecolor=\"black\")\n",
    "\n",
    "    ax.set_xticklabels(metadata2_cats, fontsize=\"xx-small\")\n",
    "    ax.set_yticklabels(metadata1_cats, fontsize=\"xx-small\")\n",
    "    ax.set_xlabel(metadata2.upper(), fontsize=\"xx-small\")\n",
    "    ax.set_ylabel(metadata1.upper(), fontsize=\"xx-small\")\n",
    "    ax.set_title(f\"Skin Tone Screentime for {metadata1.upper()} x {metadata2.upper()}\", fontsize=\"xx-small\")\n",
    "    filename = f\"plots/rp_x_metadata_x_metadata/skin_tone_x_{metadata1}_x_{metadata2}.png\"\n",
    "    figure = ax.get_figure()\n",
    "    figure.savefig(filename, dpi=300, bbox_inches=\"tight\")\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "skin_tone_heatmap(data_df, rp_data2, \"country\", \"brand\")\n",
    "skin_tone_heatmap(data_df, rp_data2, \"product type\", \"brand\")\n",
    "skin_tone_heatmap(data_df, rp_data2, \"product type\", \"country\")\n",
    "skin_tone_heatmap(data_df, rp_data2, \"product target gender\", \"brand\")\n",
    "skin_tone_heatmap(data_df, rp_data2, \"product target gender\", \"country\")\n",
    "skin_tone_heatmap(data_df, rp_data2, \"product target gender\", \"product type\")"
   ]
  },
  {
   "attachments": {},
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Target Gender x Metadata x Metadata HeatMap"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "def target_gender_metadata_x_metadata_heatmap(data_df, metadata1, metadata2):\n",
    "    seaborn.set_style(\"dark\")\n",
    "    metadata1_cats_dist = collections.Counter(data_df[metadata1])\n",
    "    metadata1_cats = filter(lambda cat: metadata1_cats_dist[cat] >= 30, metadata1_cats_dist.keys())\n",
    "    metadata1_cats = sorted(metadata1_cats, key=lambda cat: metadata1_cats_dist[cat], reverse=True)\n",
    "    metadata2_cats_dist = collections.Counter(data_df[metadata2])\n",
    "    metadata2_cats = filter(lambda cat: metadata2_cats_dist[cat] >= 30, metadata2_cats_dist.keys())\n",
    "    metadata2_cats = sorted(metadata2_cats, key=lambda cat: metadata2_cats_dist[cat], reverse=True)\n",
    "    target_gender_cats = sorted(data_df[\"product target gender\"].unique())\n",
    "    value_arr = np.zeros((len(metadata1_cats), len(metadata2_cats), len(target_gender_cats)), dtype=int)\n",
    "    for (metadata1_cat, metadata2_cat, target_gender_cat), df in data_df.groupby([metadata1, metadata2, \n",
    "                                                                                  \"product target gender\"]):\n",
    "        if metadata1_cat in metadata1_cats and metadata2_cat in metadata2_cats:\n",
    "            i, j, k = (metadata1_cats.index(metadata1_cat), metadata2_cats.index(metadata2_cat), \n",
    "                       target_gender_cats.index(target_gender_cat))\n",
    "            value_arr[i, j, k] = len(df)\n",
    "    count_arr = value_arr.sum(axis=2)\n",
    "    arr = value_arr/(1e-8 + value_arr.sum(axis=2, keepdims=True))\n",
    "    arr[value_arr == 0] = 0\n",
    "    arr[count_arr < 30] = np.nan\n",
    "    cmap_names = [\"Reds\", \"Blues\", \"Greens\", \"Wistia\"]\n",
    "    plt.figure(figsize=(50, 20))\n",
    "    for i, target_gender_cat in enumerate(target_gender_cats):\n",
    "        ax = plt.subplot(2, 2, i + 1)\n",
    "        heatmap_arr = arr[:, :, i]\n",
    "        cmap = matplotlib.colormaps[cmap_names[i]](np.linspace(0.2, 0.9, 100))\n",
    "        cmap = matplotlib.colors.ListedColormap(cmap)\n",
    "        seaborn.heatmap(100*heatmap_arr, vmin=0, vmax=100, cmap=cmap, annot=True, fmt=\".2f\", linewidths=2, \n",
    "                        linecolor=\"black\", xticklabels=metadata2_cats, yticklabels=metadata1_cats, ax=ax, cbar=True,\n",
    "                        annot_kws={\"fontsize\": \"x-small\"}, cbar_kws={\"shrink\": 0.75, \"fraction\": 0.1, \"pad\": 0.05})\n",
    "        ax.set_xticklabels(metadata2_cats, fontsize=\"xx-small\")\n",
    "        ax.set_yticklabels(metadata1_cats, fontsize=\"xx-small\")\n",
    "        ax.set_xlabel(metadata2.upper(), fontsize=\"xx-small\")\n",
    "        ax.set_ylabel(metadata1.upper(), fontsize=\"xx-small\")\n",
    "        ax.set_title(f\"product target gender = {target_gender_cat.upper()} for {metadata1.lower()} x {metadata2.lower()}\")\n",
    "    plt.subplots_adjust(hspace=0.35, wspace=0.02)\n",
    "    filename = f\"plots/product_target_gender_metadata_x_metadata/product_target_gender_x_{metadata1}_x_{metadata2}\"\n",
    "    plt.savefig(filename, dpi=300, bbox_inches=\"tight\")\n",
    "    plt.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "target_gender_metadata_x_metadata_heatmap(data_df, \"product type\", \"brand\")\n",
    "target_gender_metadata_x_metadata_heatmap(data_df, \"product type\", \"country\")\n",
    "target_gender_metadata_x_metadata_heatmap(data_df, \"country\", \"brand\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "def metadata_x_metadata_count(data_df, metadata1, metadata2):\n",
    "    metadata1_cats = data_df[metadata1].unique()\n",
    "    metadata2_cats = data_df[metadata2].unique()\n",
    "    count = np.zeros((len(metadata1_cats), len(metadata2_cats)), dtype=int)\n",
    "    for (metadata1_cat, metadata2_cat), gdf in data_df.groupby([metadata1, metadata2]):\n",
    "        i = (metadata1_cats == metadata1_cat).nonzero()[0].item()\n",
    "        j = (metadata2_cats == metadata2_cat).nonzero()[0].item()\n",
    "        count[i, j] = len(gdf)\n",
    "    count_df = pd.DataFrame(count, index=metadata1_cats, columns=metadata2_cats)\n",
    "    display(count_df)"
   ]
  },
  {
   "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>ARMANI</th>\n",
       "      <th>GARNIER</th>\n",
       "      <th>LANCOME</th>\n",
       "      <th>LRP</th>\n",
       "      <th>MNY</th>\n",
       "      <th>OAP</th>\n",
       "      <th>VICHY</th>\n",
       "      <th>YSL</th>\n",
       "      <th>CERAVE</th>\n",
       "      <th>KERASTASE</th>\n",
       "      <th>LOP</th>\n",
       "      <th>KIEHLS</th>\n",
       "      <th>MAYBELLINE</th>\n",
       "      <th>NYX</th>\n",
       "      <th>PRADA</th>\n",
       "      <th>REDKEN</th>\n",
       "      <th>BIOTHERM</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>fragrance</th>\n",
       "      <td>268</td>\n",
       "      <td>3</td>\n",
       "      <td>166</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>160</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>84</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>haircare</th>\n",
       "      <td>0</td>\n",
       "      <td>844</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>290</td>\n",
       "      <td>34</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>333</td>\n",
       "      <td>109</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>30</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>makeup</th>\n",
       "      <td>65</td>\n",
       "      <td>536</td>\n",
       "      <td>174</td>\n",
       "      <td>73</td>\n",
       "      <td>374</td>\n",
       "      <td>258</td>\n",
       "      <td>81</td>\n",
       "      <td>448</td>\n",
       "      <td>20</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>200</td>\n",
       "      <td>318</td>\n",
       "      <td>286</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>61</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>skincare</th>\n",
       "      <td>4</td>\n",
       "      <td>442</td>\n",
       "      <td>164</td>\n",
       "      <td>558</td>\n",
       "      <td>34</td>\n",
       "      <td>331</td>\n",
       "      <td>396</td>\n",
       "      <td>5</td>\n",
       "      <td>219</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>229</td>\n",
       "      <td>28</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>19</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           ARMANI  GARNIER  LANCOME  LRP  MNY  OAP  VICHY  YSL  CERAVE  \\\n",
       "fragrance     268        3      166    0    0    1      0  160       0   \n",
       "haircare        0      844        0    0    0  290     34    0       0   \n",
       "makeup         65      536      174   73  374  258     81  448      20   \n",
       "skincare        4      442      164  558   34  331    396    5     219   \n",
       "\n",
       "           KERASTASE  LOP  KIEHLS  MAYBELLINE  NYX  PRADA  REDKEN  BIOTHERM  \n",
       "fragrance          0    0       0           0    0     84       0         0  \n",
       "haircare         333  109       0           0    0      0      30         0  \n",
       "makeup             2    1     200         318  286      0       0        61  \n",
       "skincare           0    0     229          28    0      0       0        19  "
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "metadata_x_metadata_count(data_df, \"product type\", \"brand\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 94,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fragrance 0.009522731724089245\n",
      "haircare 5.218098084605763e-60\n",
      "makeup 4.0421912770118626e-19\n",
      "skincare 3.7931340619179835e-14\n"
     ]
    }
   ],
   "source": [
    "for product_type in data_df[\"product type\"].unique():\n",
    "    arr = rp_data2[data_df[\"product type\"] == product_type].sum(axis=(1, 2))\n",
    "    arr = (arr[:, 1] - arr[:, 2])/arr.sum(axis=1)\n",
    "    _, pvalue = scipy.stats.ttest_1samp(arr, 0, alternative=\"two-sided\")\n",
    "    print(product_type, pvalue)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 95,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "female 2.7636716762263127e-08\n",
      "male 8.426556860259266e-05\n",
      "unspecified 3.082001238058868e-45\n",
      "unisex 1.797347954479892e-05\n"
     ]
    }
   ],
   "source": [
    "for product_type in data_df[\"product target gender\"].unique():\n",
    "    arr = rp_data2[data_df[\"product target gender\"] == product_type].sum(axis=(1, 2))\n",
    "    arr = (arr[:, 1] - arr[:, 2])/arr.sum(axis=1)\n",
    "    _, pvalue = scipy.stats.ttest_1samp(arr, 0, alternative=\"two-sided\")\n",
    "    print(product_type, pvalue)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.05156743728834456"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "arr = rp_data2[(data_df[\"brand\"] == \"ARMANI\") & (data_df[\"product type\"] == \"makeup\")].sum(axis=(1, 2))\n",
    "arr = arr[:, 0]/arr.sum(axis=1)\n",
    "_, pvalue = scipy.stats.ttest_1samp(arr, 0.5, alternative=\"greater\")\n",
    "pvalue"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.11.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}