data_preprocessing.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "BeBJWOaYHMbZ",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Qwnv3DBTCZZP",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "# Preprocessing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ZjAmPTP1Mg6E",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "## Installation and import of the needed libraries\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "oYnJw3nRtWjx",
    "outputId": "2771b510-ae60-4858-d747-7c73655d7b9e",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Currently not running in a Colab environment\n"
     ]
    }
   ],
   "source": [
    "COLAB_ENV = (\n",
    "    \"google.colab\" in str(get_ipython())\n",
    "    if hasattr(__builtins__, \"__IPYTHON__\")\n",
    "    else False\n",
    ")\n",
    "print(f\"Currently {'not ' if not COLAB_ENV else ''}running in a Colab environment\")\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "PlR3PcJbC84d",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "### Installation (mandatory on Colab)\n",
    "\n",
    "First of all we clone the Github repository in order to have all the needed files to perform our analysis.\n",
    "\n",
    "This is needed only if the current notebook isn't executed directly from the repository folder (e.g. Google Colab)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "bXnBmuobMi-Q",
    "outputId": "bc57fb44-aadb-4abb-cac6-f61156d9f607",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "if COLAB_ENV:\n",
    "    !mkdir my_repo\n",
    "    !git clone https://github.com/Valendrew/counterfactual-explanations.git my_repo\n",
    "    %cd my_repo\n",
    "    !pip install -r requirements.txt -q"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YCxP1Q5-HMbi",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "### Import and configuration setup\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "QZR9pR0LSjbE",
    "outputId": "fbdbf73e-53f4-429b-d089-1f3bb474b31f",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "# python modules\n",
    "import configparser\n",
    "import re\n",
    "\n",
    "# 3rd party modules\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.linear_model import LinearRegression, HuberRegressor\n",
    "from sklearn.model_selection import cross_val_score\n",
    "from sklearn.model_selection import RandomizedSearchCV\n",
    "\n",
    "# user modules\n",
    "from utils import util_data, util_preprocess, util_plot, util_models"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "zmh5RLDwSRwx",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "In order to manage the randomness we wrote the following cell to make the experiments reproducible, check the following [link](https://scikit-learn.org/stable/common_pitfalls.html#controlling-randomness).\n",
    "\n",
    "A numpy Random Generator object is constructed for controlling randomness: necessary for robust results in estimator performance with cross-validation.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "id": "5I7sSu9aSY3D",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "def set_default_randomness(seed: int):\n",
    "    rng = np.random.RandomState(seed)\n",
    "    return rng\n",
    "\n",
    "\n",
    "SEED = 42\n",
    "rng = set_default_randomness(SEED)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "CcWgPz39UFd_",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Moreover we import the configuration file to have all the parameters along the notebook.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "P0trduHkUG3S",
    "outputId": "ad4aa90e-7d18-4afb-9d60-2998fc599a4a",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['config.ini']"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "config = configparser.ConfigParser()\n",
    "config.read(\"config.ini\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "-8tIL4WpMteF",
    "outputId": "e836dd86-e039-4b27-ccfa-ba3def199e2b",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File gsm.csv already exists. Skip download.\n"
     ]
    }
   ],
   "source": [
    "ds_conf = config[\"GSMARENA_DS\"]\n",
    "# Download the training and test datasets\n",
    "download_helper = util_data.DownloadHelper(\n",
    "    ds_conf[\"url\"], ds_conf[\"name\"], ds_conf[\"mode\"], quiet=False\n",
    ")\n",
    "\n",
    "download_helper.download()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "id": "9RjANCWpOq90",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "df_raw = download_helper.read_csv(low_memory=False, sep=\",\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "tETAOV0DJV6d",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "## Exploration of scraped data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "For the project we use the GSMArena Mobile Phone Devices dataset ([Kaggle link](https://www.kaggle.com/datasets/msainani/gsmarena-mobile-devices)).\n",
    "\n",
    "The dataset contains information about different characteristics of the smartphones (e.g. RAM, cores, ...) and the target variable is the price range of the device. The data has been scraped from [GSMArena](gsmarena.com), a well-known website that provides information and specification about smartphones and other electronic devices."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Dataset information"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "8DoMLVVA4EHS",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The dataset has 10679 samples with 86 features. The features are a mix of numerical and categorical data, but some of them are redundant or not useful for our analysis (e.g. battery tests, network bands supported, ...). \n",
    "\n",
    "Moreover, some of the features are not well formatted (e.g. the price is a string with the currency symbol), thus we need to perform some preprocessing to make the dataset usable for our analysis and to extract the most useful features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 393
    },
    "id": "GKfO-zuZ0ykr",
    "outputId": "a73b8bea-bec0-4303-bcb6-4de338807899",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>oem</th>\n",
       "      <th>model</th>\n",
       "      <th>network_technology</th>\n",
       "      <th>network_2g_bands</th>\n",
       "      <th>network_gprs</th>\n",
       "      <th>network_edge</th>\n",
       "      <th>launch_announced</th>\n",
       "      <th>launch_status</th>\n",
       "      <th>body_dimensions</th>\n",
       "      <th>body_weight</th>\n",
       "      <th>...</th>\n",
       "      <th>main_camera_dual_or_triple</th>\n",
       "      <th>battery_music_play</th>\n",
       "      <th>selfie_camera_triple</th>\n",
       "      <th>main_camera_v1</th>\n",
       "      <th>selfie_camera</th>\n",
       "      <th>camera</th>\n",
       "      <th>main_camera</th>\n",
       "      <th>network</th>\n",
       "      <th>battery_talk_time</th>\n",
       "      <th>battery_stand.by</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>3158</th>\n",
       "      <td>Celkon</td>\n",
       "      <td>Millennia Everest</td>\n",
       "      <td>GSM / HSPA</td>\n",
       "      <td>GSM 900 / 1800 - SIM 1 &amp; SIM 2</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>2015, August</td>\n",
       "      <td>Available. Released 2015, August</td>\n",
       "      <td>157 x 78 x 9.1 mm (6.18 x 3.07 x 0.36 in)</td>\n",
       "      <td>-</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3657</th>\n",
       "      <td>Gionee</td>\n",
       "      <td>Elife E6</td>\n",
       "      <td>GSM / HSPA</td>\n",
       "      <td>GSM 850 / 900 / 1800 / 1900</td>\n",
       "      <td>Class 12</td>\n",
       "      <td>Class 12</td>\n",
       "      <td>2013. Released 2013</td>\n",
       "      <td>Discontinued</td>\n",
       "      <td>142.5 x 69.4 x 7.9 mm (5.61 x 2.73 x 0.31 in)</td>\n",
       "      <td>-</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4403</th>\n",
       "      <td>i-mobile</td>\n",
       "      <td>903</td>\n",
       "      <td>GSM</td>\n",
       "      <td>GSM 900 / 1800 / 1900</td>\n",
       "      <td>Class 10</td>\n",
       "      <td>No</td>\n",
       "      <td>2007, May. Released 2007, May</td>\n",
       "      <td>Discontinued</td>\n",
       "      <td>106 x 47 x 17 mm (4.17 x 1.85 x 0.67 in)</td>\n",
       "      <td>88 g (3.10 oz)</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>No</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>Up to 2 h</td>\n",
       "      <td>Up to 200 h</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3 rows × 86 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "           oem              model network_technology  \\\n",
       "3158    Celkon  Millennia Everest         GSM / HSPA   \n",
       "3657    Gionee           Elife E6         GSM / HSPA   \n",
       "4403  i-mobile                903                GSM   \n",
       "\n",
       "                    network_2g_bands network_gprs network_edge  \\\n",
       "3158  GSM 900 / 1800 - SIM 1 & SIM 2          Yes          Yes   \n",
       "3657     GSM 850 / 900 / 1800 / 1900     Class 12     Class 12   \n",
       "4403           GSM 900 / 1800 / 1900     Class 10           No   \n",
       "\n",
       "                   launch_announced                     launch_status  \\\n",
       "3158                   2015, August  Available. Released 2015, August   \n",
       "3657            2013. Released 2013                      Discontinued   \n",
       "4403  2007, May. Released 2007, May                      Discontinued   \n",
       "\n",
       "                                    body_dimensions     body_weight  ...  \\\n",
       "3158      157 x 78 x 9.1 mm (6.18 x 3.07 x 0.36 in)               -  ...   \n",
       "3657  142.5 x 69.4 x 7.9 mm (5.61 x 2.73 x 0.31 in)               -  ...   \n",
       "4403       106 x 47 x 17 mm (4.17 x 1.85 x 0.67 in)  88 g (3.10 oz)  ...   \n",
       "\n",
       "     main_camera_dual_or_triple battery_music_play selfie_camera_triple  \\\n",
       "3158                        NaN                NaN                  NaN   \n",
       "3657                        NaN                NaN                  NaN   \n",
       "4403                        NaN                NaN                  NaN   \n",
       "\n",
       "     main_camera_v1 selfie_camera camera main_camera network  \\\n",
       "3158            NaN           NaN    NaN         NaN     NaN   \n",
       "3657            NaN           NaN    NaN         NaN     NaN   \n",
       "4403            NaN            No    NaN         NaN     NaN   \n",
       "\n",
       "     battery_talk_time battery_stand.by  \n",
       "3158               NaN              NaN  \n",
       "3657               NaN              NaN  \n",
       "4403         Up to 2 h      Up to 200 h  \n",
       "\n",
       "[3 rows x 86 columns]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_raw.sample(3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "atmzrDaoHMbl",
    "outputId": "df887a31-f069-48f3-d35f-3512cc2bd137",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of rows: 10668\n",
      "Number of columns: 86\n"
     ]
    }
   ],
   "source": [
    "print(f\"Number of rows: {df_raw.shape[0]}\")\n",
    "print(f\"Number of columns: {df_raw.shape[1]}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "q583kKVEHMbl",
    "outputId": "66b6e375-692f-4c1a-8fbd-767d84af39f9",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Columns:\n",
      "Index(['oem', 'model', 'network_technology', 'network_2g_bands',\n",
      "       'network_gprs', 'network_edge', 'launch_announced', 'launch_status',\n",
      "       'body_dimensions', 'body_weight', 'body_sim', 'display_type',\n",
      "       'display_size', 'display_resolution', 'display', 'memory_card_slot',\n",
      "       'memory_phonebook', 'memory_call_records', 'sound_loudspeaker',\n",
      "       'sound_alert_types', 'sound_3.5mm_jack', 'comms_wlan',\n",
      "       'comms_bluetooth', 'comms_gps', 'comms_radio', 'comms_usb',\n",
      "       'features_sensors', 'features_messaging', 'features_browser',\n",
      "       'features_clock', 'features_alarm', 'features_games', 'features_java',\n",
      "       'features', 'misc_colors', 'network_3g_bands', 'network_speed',\n",
      "       'platform_os', 'platform_chipset', 'platform_cpu', 'platform_gpu',\n",
      "       'memory_internal', 'main_camera_single', 'main_camera_video',\n",
      "       'misc_price', 'main_camera_features', 'body', 'network_4g_bands',\n",
      "       'body_build', 'display_protection', 'memory', 'main_camera_dual',\n",
      "       'selfie_camera_dual', 'selfie_camera_features', 'selfie_camera_video',\n",
      "       'comms_nfc', 'battery_charging', 'misc_models', 'tests_performance',\n",
      "       'tests_camera', 'tests_loudspeaker', 'tests_audio_quality',\n",
      "       'tests_battery_life', 'tests_display', 'selfie_camera_single',\n",
      "       'comms_infrared_port', 'network_5g_bands', 'main_camera_quad',\n",
      "       'main_camera_triple', 'sound', 'misc_sar_eu', 'main_camera_five',\n",
      "       'features_languages', 'body_keyboard', 'misc_sar', 'battery',\n",
      "       'main_camera_dual_or_triple', 'battery_music_play',\n",
      "       'selfie_camera_triple', 'main_camera_v1', 'selfie_camera', 'camera',\n",
      "       'main_camera', 'network', 'battery_talk_time', 'battery_stand.by'],\n",
      "      dtype='object')\n"
     ]
    }
   ],
   "source": [
    "print(f\"Columns:\\n{df_raw.columns}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "fIjcQFmL0Lbb",
    "outputId": "938c0afb-54c3-477e-8286-a35a028b3e1a",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "object    86\n",
       "dtype: int64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_raw.dtypes.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_cVSjvYxtSeN",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "### Explore relevant features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "u5xSZZn7tSeN",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- _misc_price_: label for the classification task\n",
    "- _launch_announced_: year of the announcement of the device (used only for filtering)\n",
    "- _network_technology_: 3G, 4G, 5G support\n",
    "- **Numerical features**: display_size, battery\n",
    "- **Binary features**: display_type, memory_card_slot, sound_loudspeaker, sound_3.5mm_jack, comms_wlan, comms_nfc, battery_charging, body\n",
    "- **Multi column features**: display_resolution, memory_internal\n",
    "- **Camera features**: main_camera, selfie_camera\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "id": "VwU2wdX6tSeO",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "df_copy = df_raw.copy()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xzp8KKLqtSeO",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "#### Explore the launch_announced column"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Nu9UdFgPtSeO",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "There are 91 values which are NaN. However, these samples belongs to two distinct categories:\n",
    "\n",
    "- 82 samples don't have any price, so without a proper label they must be dropped.\n",
    "- 9 samples have a price, but carefully looking at the \"launch_status\", only two of them have been \"discontinued\", while the remaining ones have been \"cancelled\"."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Nonetheless, these 91 samples can be easily dropped, because only for 2 of them the launch year could be found, but a quick online search doesn't provide any results."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "def extract_year(df: pd.DataFrame, feat_col: str):\n",
    "    # pattern to extract year\n",
    "    pat_year = r\"([\\d]{4})\"\n",
    "    # year series\n",
    "    ser_year = df_copy[feat_col]\n",
    "    \n",
    "    ser_year = util_preprocess.DataTransformer.process_float_feature(ser_year, pat_year)\n",
    "    return ser_year"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of missing values: 91\n"
     ]
    }
   ],
   "source": [
    "ser_year = extract_year(df_copy, \"launch_announced\")\n",
    "print(f\"Number of missing values: {ser_year.isna().sum()}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "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>oem</th>\n",
       "      <th>model</th>\n",
       "      <th>launch_announced</th>\n",
       "      <th>launch_status</th>\n",
       "      <th>misc_price</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>853</th>\n",
       "      <td>Xiaomi</td>\n",
       "      <td>Mi Max 4</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 300 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>854</th>\n",
       "      <td>Xiaomi</td>\n",
       "      <td>Mi Max 4 Pro</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 350 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1642</th>\n",
       "      <td>Coolpad</td>\n",
       "      <td>3632</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 100 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2610</th>\n",
       "      <td>Asus</td>\n",
       "      <td>Z801</td>\n",
       "      <td>V1</td>\n",
       "      <td>Discontinued</td>\n",
       "      <td>About 260 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4915</th>\n",
       "      <td>Lenovo</td>\n",
       "      <td>ideapad</td>\n",
       "      <td>V1</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 130 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5584</th>\n",
       "      <td>Meizu</td>\n",
       "      <td>PRO 5 mini</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 360 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8975</th>\n",
       "      <td>Samsung</td>\n",
       "      <td>Galaxy On5 (2016)</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 220 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10059</th>\n",
       "      <td>Vodafone</td>\n",
       "      <td>830i</td>\n",
       "      <td>V1</td>\n",
       "      <td>Discontinued</td>\n",
       "      <td>About 170 EUR</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10233</th>\n",
       "      <td>Xiaomi</td>\n",
       "      <td>Redmi Pro 2</td>\n",
       "      <td>Not announced yet</td>\n",
       "      <td>Cancelled</td>\n",
       "      <td>About 210 EUR</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            oem              model   launch_announced launch_status  \\\n",
       "853      Xiaomi           Mi Max 4  Not announced yet     Cancelled   \n",
       "854      Xiaomi       Mi Max 4 Pro  Not announced yet     Cancelled   \n",
       "1642    Coolpad               3632  Not announced yet     Cancelled   \n",
       "2610       Asus               Z801                 V1  Discontinued   \n",
       "4915     Lenovo            ideapad                 V1     Cancelled   \n",
       "5584      Meizu         PRO 5 mini  Not announced yet     Cancelled   \n",
       "8975    Samsung  Galaxy On5 (2016)  Not announced yet     Cancelled   \n",
       "10059  Vodafone               830i                 V1  Discontinued   \n",
       "10233    Xiaomi        Redmi Pro 2  Not announced yet     Cancelled   \n",
       "\n",
       "          misc_price  \n",
       "853    About 300 EUR  \n",
       "854    About 350 EUR  \n",
       "1642   About 100 EUR  \n",
       "2610   About 260 EUR  \n",
       "4915   About 130 EUR  \n",
       "5584   About 360 EUR  \n",
       "8975   About 220 EUR  \n",
       "10059  About 170 EUR  \n",
       "10233  About 210 EUR  "
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols = [\"oem\", \"model\", \"launch_announced\", \"launch_status\", \"misc_price\"]\n",
    "df_copy.loc[ser_year.isna(), cols].dropna(subset=\"misc_price\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3bAoawxWHMbm",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Explore the misc_price column"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1JpoJDCPtSeP",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "There are only 6527 valid price rows, while the remaining ones are NaN values, so they must be dropped.\n",
    "\n",
    "For most of the models the currency available is the *EUR*, but for the remaining ones (*USD*, *INR*, *GBP*) a suitable conversion rate should be found to estimate the actual price in EUR, otherwise the price could be misleading for the classification task given that the price range for most of the samples is in euros.\n",
    "\n",
    "However, finding a suitable conversion rate for each region market is non trivial (e.g. difference [purchasing power parities](https://data.oecd.org/conversion/purchasing-power-parities-ppp.htm)), so we decided to compute conversion rates for all smartphones wherever the currency is both in *EUR* and *USD/INR/GBP*, then training a model to predict the price in EUR and compare the errors with two other approaches:\n",
    "- An online conversion rate, retrieved from [here](https://www.ecb.europa.eu/stats/policy_and_exchange_rates/euro_reference_exchange_rates/html/index.en.html) in the span of the sample years (usually 1/2 years).\n",
    "- The median conversion rate among the selected samples."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "FfAL2as9HMbm",
    "outputId": "09b7908d-79cb-4ac4-9d23-3e7a3b586348",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of valid 'misc_price' rows: 6527\n"
     ]
    }
   ],
   "source": [
    "misc_price = df_copy[\"misc_price\"]\n",
    "to_replace = {\n",
    "    \"<e2><80><89>\": \" \",\n",
    "    \"<e2><82><ac>\": \"EUR\",\n",
    "    \"<e2><82><B9>\": \"INR\",\n",
    "    \"<c2><a3>\": \"GBP\",\n",
    "    \"€\": \"EUR\",\n",
    "    \"\\$\": \"USD\",\n",
    "    \"£\": \"GBP\",\n",
    "    \"\" \"\\u2009\": \" \",\n",
    "    \"(,)(?=[\\d]{3})\": \"\",\n",
    "}\n",
    "for k, v in to_replace.items():\n",
    "    misc_price = misc_price.str.replace(k, v, regex=True, flags=re.IGNORECASE)\n",
    "\n",
    "print(f\"Number of valid 'misc_price' rows: {misc_price.count()}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Conversion rate between USD and EUR."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The conversion rate retrieved from the European Central Bank is *0.8694*, from which the VAT is [multiplied](https://www.adventurespedia.com/post/why-doesnt-the-u-s-include-tax-in-prices/#:~:text=The%20U.S.%20does%20not%20include,to%20the%20price%20in%20America.) to obtain the actual conversion rate, that is *1.0606*, while the median conversion rate is *1.0757*.\n",
    "\n",
    "To estimate the conversion rate between USD and EUR we can use a linear regression model, but given the fact that some prices are outliers (e.g. rate higher than 1.70), we need a more robust model, so we use the Huber regressor."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "metadata": {
    "id": "7CThoe0htSeT",
    "outputId": "4a990195-281c-4066-8133-b8173c5d485c",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of prices: 164\n"
     ]
    }
   ],
   "source": [
    "curr = \"USD\"\n",
    "prices_ext, conversion_rate = util_preprocess.extract_prices_rate(curr, misc_price)\n",
    "\n",
    "# add the year to the conversion rate\n",
    "conv_rate_year = pd.concat([conversion_rate, ser_year.loc[conversion_rate.index]], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 103,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Year coverage: [2018. 2019. 2020.]\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYYAAAEiCAYAAAD9DXUdAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAcMklEQVR4nO3dcXDT9eH/8VdKQ6Dagigk7VGxsgLDTg9FS0UFlcah42TcTV03Dj11OtCt6+0q2NuZftUCva1jWx07NgbdHxW3YdU7FJq7SZlWvKJ1c+gYzoqdtHIotLXFNNL37w9/zZc3TUvTJk2y7/Nxl+t93p93Pnn1w7t9kaT91GGMMQIA4P9LiXcAAEBioRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAAJbUeAc4W19fn44ePar09HQ5HI54xwGA/wrGGHV1dSkrK0spKUM/J0i4Yjh69Kiys7PjHQMA/iu1trZq+vTpQ85JuGJIT0+X9GX4jIyMAfuDwaDq6+vl9XrldDrHOl7Eki2vROaxkmyZky2vROYzdXZ2Kjs7O/Q9digRFYPP51N5ebk15na71d7eLunLpyrl5eXasmWLTpw4ofz8fD311FO67LLLhv0Y/S8fZWRkDFoMaWlpysjISIp/6GTLK5F5rCRb5mTLK5E5nOG8RB/xm8+XXXaZ2traQre33347tK+yslJVVVWqrq5WU1OTPB6PCgsL1dXVFenDAADiJOJiSE1NlcfjCd2mTp0q6ctnC5s2bVJZWZlWrFihvLw81dTUqKenR7W1tVEPDgCIjYiL4fDhw8rKylJOTo7uuusuvf/++5KklpYWtbe3y+v1hua6XC4tWrRIjY2N0UsMAIipiN5jyM/P1x/+8AfNmjVLH3/8sZ544glde+21OnjwYOh9Brfbbd3H7XbryJEjgx4zEAgoEAiEtjs7OyV9+TpbMBgcML9/LNy+RJRseSUyj5Vky5xseSUyhzvucDhG84d6uru7NXPmTJWWlmrBggVauHChjh49qszMzNCc+++/X62trdq9e3fYY4R7Q1uSamtrlZaWNtJoAIAz9PT0qKioSB0dHWF/sOdMo/px1fPOO09f+9rXdPjwYS1fvlyS1N7ebhXDsWPHBjyLONO6detUUlIS2u7/kSqv1zvoTyX5/X4VFhYmxU8ZJFteicxjJdkyJ1teicxn6n81ZjhGVQyBQEDvvvuurr/+euXk5Mjj8cjv92vevHmSpN7eXjU0NGjjxo2DHsPlcsnlcg0YdzqdQ56Uc+1PNMmWVyLzWEm2zMmWVyJz//GGK6Ji+PGPf6xly5bp4osv1rFjx/TEE0+os7NTq1atksPhUHFxsSoqKpSbm6vc3FxVVFQoLS1NRUVFEX8SAID4iKgY/vOf/+jb3/62jh8/rqlTp2rBggXav3+/ZsyYIUkqLS3VqVOntHr16tAvuNXX1w/rN+0AAIkhomLYsWPHkPsdDod8Pp98Pt9oMiEJXbJ2V1wf//Dj3nNPAjAsXHYbAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAFooBAGChGAAAltR4B0D0XLJ2V1SO4xpnVHmNlOfbo8BpR1SOCSB58IwBAGChGAAAFooBAGChGAAAllEVw/r16+VwOFRcXBwaM8bI5/MpKytLEydO1OLFi3Xw4MHR5gQAjJERF0NTU5O2bNmiyy+/3BqvrKxUVVWVqqur1dTUJI/Ho8LCQnV1dY06LAAg9kZUDJ999pm+853v6Le//a0uuOCC0LgxRps2bVJZWZlWrFihvLw81dTUqKenR7W1tVELDQCInREVw5o1a3TbbbdpyZIl1nhLS4va29vl9XpDYy6XS4sWLVJjY+PokgIAxkTEv+C2Y8cOvfnmm2pqahqwr729XZLkdrutcbfbrSNHjoQ9XiAQUCAQCG13dnZKkoLBoILB4ID5/WPh9iWisczrGmeic5wUY31MBsm2LqTky5xseSUyhzvucERUDK2trfrhD3+o+vp6TZgwYdB5Dof927LGmAFj/davX6/y8vIB4/X19UpLSxv0Mfx+/zBTJ4axyFt5TXSP9/j8vugeMIb6z2+yrQsp+TInW16JzJLU09Mz7LkOY8yw/1v43HPP6Zvf/KbGjRsXGjt9+rQcDodSUlJ06NAhfeUrX9Gbb76pefPmhebcfvvtmjx5smpqagYcM9wzhuzsbB0/flwZGRkD5geDQfn9fhUWFsrpdA77E42Xscyb59sTleO4Uowen9+nnxxIUaAvOS6J0Vx2U1KtC4m1PBbI/L86Ozt10UUXqaOjI+z31jNF9Izh5ptv1ttvv22N3XPPPZozZ44eeeQRXXrppfJ4PPL7/aFi6O3tVUNDgzZu3Bj2mC6XSy6Xa8C40+kc8qSca3+iGYu80b6uUaDPkTTXSuo/t8m2LqTky5xseSUy9x9vuCIqhvT0dOXl5Vlj5513ni688MLQeHFxsSoqKpSbm6vc3FxVVFQoLS1NRUVFkTwUACBOon511dLSUp06dUqrV6/WiRMnlJ+fr/r6eqWnp0f7oQAAMTDqYti7d6+17XA45PP55PP5RntoAEAccK0kAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAICFYgAAWCgGAIAlomLYvHmzLr/8cmVkZCgjI0MFBQV66aWXQvuNMfL5fMrKytLEiRO1ePFiHTx4MOqhAQCxE1ExTJ8+XRs2bNCBAwd04MAB3XTTTbr99ttD3/wrKytVVVWl6upqNTU1yePxqLCwUF1dXTEJDwCIvoiKYdmyZbr11ls1a9YszZo1S08++aTOP/987d+/X8YYbdq0SWVlZVqxYoXy8vJUU1Ojnp4e1dbWxio/ACDKRvwew+nTp7Vjxw51d3eroKBALS0tam9vl9frDc1xuVxatGiRGhsboxIWABB7qZHe4e2331ZBQYE+//xznX/++aqrq9PcuXND3/zdbrc13+1268iRI4MeLxAIKBAIhLY7OzslScFgUMFgcMD8/rFw+xLRWOZ1jTPROU6KsT4mg2RbF1LyZU62vBKZwx13OBzGmIi++nt7e/Xhhx/q5MmT2rlzp373u9+poaFBJ0+e1MKFC3X06FFlZmaG5t9///1qbW3V7t27wx7P5/OpvLx8wHhtba3S0tIiiQYAGERPT4+KiorU0dGhjIyMIedGXAxnW7JkiWbOnKlHHnlEM2fO1Jtvvql58+aF9t9+++2aPHmyampqwt4/3DOG7OxsHT9+PGz4YDAov9+vwsJCOZ3O0UQfE2OZN8+3JyrHcaUYPT6/Tz85kKJAnyMqx4y15rKbkmpdSKzlsUDm/9XZ2amLLrpoWMUQ8UtJZzPGKBAIKCcnRx6PR36/P1QMvb29amho0MaNGwe9v8vlksvlGjDudDqHPCnn2p9oxiJv4HR0v4kH+hxRP2as9J/bZFsXUvJlTra8Epn7jzdcERXDo48+qqVLlyo7O1tdXV3asWOH9u7dq927d8vhcKi4uFgVFRXKzc1Vbm6uKioqlJaWpqKioog/CQBAfERUDB9//LFWrlyptrY2TZo0SZdffrl2796twsJCSVJpaalOnTql1atX68SJE8rPz1d9fb3S09NjEj4RXbJ2l7XtGmdUec2XL/Mky/++AfzfFlExbN26dcj9DodDPp9PPp9vNJkAAHHEtZIAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAABaKAQBgoRgAAJaIimH9+vW6+uqrlZ6ermnTpmn58uU6dOiQNccYI5/Pp6ysLE2cOFGLFy/WwYMHoxoaABA7ERVDQ0OD1qxZo/3798vv9+uLL76Q1+tVd3d3aE5lZaWqqqpUXV2tpqYmeTweFRYWqqurK+rhAQDRlxrJ5N27d1vb27Zt07Rp0/TGG2/ohhtukDFGmzZtUllZmVasWCFJqqmpkdvtVm1trR544IHoJQcAxMSo3mPo6OiQJE2ZMkWS1NLSovb2dnm93tAcl8ulRYsWqbGxcTQPBQAYIxE9YziTMUYlJSW67rrrlJeXJ0lqb2+XJLndbmuu2+3WkSNHwh4nEAgoEAiEtjs7OyVJwWBQwWBwwPz+sXD7EoFrnLG3U4z1MRkkY+ZEXxfhJFvmZMsrkTnccYfDYYwZ0Vf/mjVrtGvXLr3yyiuaPn26JKmxsVELFy7U0aNHlZmZGZp7//33q7W1dcBLUZLk8/lUXl4+YLy2tlZpaWkjiQYAOEtPT4+KiorU0dGhjIyMIeeO6BnDww8/rBdeeEH79u0LlYIkeTweSV8+czizGI4dOzbgWUS/devWqaSkJLTd2dmp7Oxseb3esOGDwaD8fr8KCwvldDpHEj+m8nx7rG1XitHj8/v0kwMpCvQ54pQqMsmYubnspoReF+Ek+lo+W7Lllch8pv5XY4YjomIwxujhhx9WXV2d9u7dq5ycHGt/Tk6OPB6P/H6/5s2bJ0nq7e1VQ0ODNm7cGPaYLpdLLpdrwLjT6RzypJxrf7wETof/Rhrocwy6L1ElU+b+tZCo62IoyZY52fJKZO4/3nBFVAxr1qxRbW2tnn/+eaWnp4feU5g0aZImTpwoh8Oh4uJiVVRUKDc3V7m5uaqoqFBaWpqKiooi+ywAAHERUTFs3rxZkrR48WJrfNu2bbr77rslSaWlpTp16pRWr16tEydOKD8/X/X19UpPT49KYABAbEX8UtK5OBwO+Xw++Xy+kWYCAMQR10oCAFgoBgCAhWIAAFgoBgCAhWIAAFgoBgCAhWIAAFgoBgCAhWIAAFgoBgCAhWIAAFgoBgCAhWIAAFhG/DefgUSS59ujymu+/DjWf1zogw23jenjAbHGMwYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgOW/8uqql6zdFe8IAJC0eMYAALBQDAAAC8UAALBQDAAAC8UAALBQDAAAC8UAALBQDAAAC8UAALBEXAz79u3TsmXLlJWVJYfDoeeee87ab4yRz+dTVlaWJk6cqMWLF+vgwYPRygsAiLGIi6G7u1tXXHGFqqurw+6vrKxUVVWVqqur1dTUJI/Ho8LCQnV1dY06LAAg9iK+VtLSpUu1dOnSsPuMMdq0aZPKysq0YsUKSVJNTY3cbrdqa2v1wAMPjC4tACDmonoRvZaWFrW3t8vr9YbGXC6XFi1apMbGxrDFEAgEFAgEQtudnZ2SpGAwqGAwOGB+/1i4faHHHGdG/DlEmyvFWB+TAZkjM9RaHM79Rnr/sZZseSUyhzvucDiMMSP+SnI4HKqrq9Py5cslSY2NjVq4cKE++ugjZWVlheZ973vf05EjR7Rnz54Bx/D5fCovLx8wXltbq7S0tJFGAwCcoaenR0VFRero6FBGRsaQc2Ny2W2Hw2FtG2MGjPVbt26dSkpKQtudnZ3Kzs6W1+sNGz4YDMrv96uwsFBOpzPsMfN8AwsoXlwpRo/P79NPDqQo0Bf+HCQaMkfmH75bRnS/4azlRJJseSUyn6n/1ZjhiGoxeDweSVJ7e7syMzND48eOHZPb7Q57H5fLJZfLNWDc6XQOeVKG2h84nXjfzAJ9joTMNRQyD89ov3jPtdYTTbLllcjcf7zhiurvMeTk5Mjj8cjv94fGent71dDQoGuvvTaaDwUAiJGInzF89tlneu+990LbLS0teuuttzRlyhRdfPHFKi4uVkVFhXJzc5Wbm6uKigqlpaWpqKgoqsEBALERcTEcOHBAN954Y2i7//2BVatWafv27SotLdWpU6e0evVqnThxQvn5+aqvr1d6enr0UgMJZKR/StY1zqjymi/fExvpy18fbLhtRPcDhhJxMSxevFhD/SCTw+GQz+eTz+cbTS4AQJxwrSQAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYKAYAgIViAABYUuMdAMDIXbJ215g9lmucUeU1Up5vjwKnHfpgw21j9thnG+7nfXbmZDBU5rE65zxjAABYKAYAgIViAABYKAYAgCVmxfDrX/9aOTk5mjBhgq666ir99a9/jdVDAQCiKCbF8Mwzz6i4uFhlZWVqbm7W9ddfr6VLl+rDDz+MxcMBAKIoJsVQVVWle++9V/fdd5+++tWvatOmTcrOztbmzZtj8XAAgCiK+u8x9Pb26o033tDatWutca/Xq8bGxgHzA4GAAoFAaLujo0OS9OmnnyoYDA6YHwwG1dPTo08++UROpzNshtQvukfzKURVap9RT0+fUoMpOt2XHD9HTeaxkWyZz877ySefxC/LML/Gk+0cS0NnHs057+rqkiQZY8492UTZRx99ZCSZV1991Rp/8sknzaxZswbMf+yxx4wkbty4ceM2BrfW1tZzfh+P2W8+Oxx20xljBoxJ0rp161RSUhLa7uvr06effqoLL7ww7PzOzk5lZ2ertbVVGRkZ0Q8eZcmWVyLzWEm2zMmWVyLzmYwx6urqUlZW1jnnRr0YLrroIo0bN07t7e3W+LFjx+R2uwfMd7lccrlc1tjkyZPP+TgZGRlJ8w8tJV9eicxjJdkyJ1teicz9Jk2aNKx5UX/zefz48brqqqvk9/utcb/fr2uvvTbaDwcAiLKYvJRUUlKilStXav78+SooKNCWLVv04Ycf6sEHH4zFwwEAoigmxXDnnXfqk08+0f/8z/+ora1NeXl5evHFFzVjxoxRH9vlcumxxx4b8PJTokq2vBKZx0qyZU62vBKZR8phzHB+dgkA8H8F10oCAFgoBgCAhWIAAFgoBgCAJe7FEMnluffu3SuHwzHg9s9//tOat3PnTs2dO1cul0tz585VXV1d3DLffffdYTNfdtlloTnbt28PO+fzzz8fddZ9+/Zp2bJlysrKksPh0HPPPXfO+zQ0NOiqq67ShAkTdOmll+o3v/nNgDmxPMeRZn722WdVWFioqVOnKiMjQwUFBdqzZ481J5bneCSZ472WI80b73UsSevXr9fVV1+t9PR0TZs2TcuXL9ehQ4fOeb94reeR5E2EtSzFuRhGennuQ4cOqa2tLXTLzc0N7Xvttdd05513auXKlfrb3/6mlStX6o477tDrr78el8y/+MUvrKytra2aMmWKvvWtb1nzMjIyrHltbW2aMGHCqPN2d3friiuuUHV19bDmt7S06NZbb9X111+v5uZmPfroo/rBD36gnTt3hubE+hxHmnnfvn0qLCzUiy++qDfeeEM33nijli1bpubmZmterM7xSDL3i9dajjRvvNex9OU3+DVr1mj//v3y+/364osv5PV61d09+AX14rmeR5I3EdayJEX9InqRuOaaa8yDDz5ojc2ZM8esXbs27PyXX37ZSDInTpwY9Jh33HGH+frXv26N3XLLLeauu+4adV5jIs98trq6OuNwOMwHH3wQGtu2bZuZNGlSVPINRZKpq6sbck5paamZM2eONfbAAw+YBQsWhLZjfY7PNJzM4cydO9eUl5eHtsfqHBszvMyJsJb7jeQcx3Md9zt27JiRZBoaGgadk0jreTh5w4nHWo7bM4b+y3N7vV5rfLDLc59p3rx5yszM1M0336yXX37Z2vfaa68NOOYtt9xyzmPGOnO/rVu3asmSJQN+2e+zzz7TjBkzNH36dH3jG98Y8D+EsTLY+Ttw4EDoMuixPMfR0NfXp66uLk2ZMsUaT5RzfKZ4reXRSoR13H+J/rP/nc+USOt5OHnPFq+1HLdiOH78uE6fPj3gwnput3vABfj6ZWZmasuWLdq5c6eeffZZzZ49WzfffLP27dsXmtPe3h7RMWOd+UxtbW166aWXdN9991njc+bM0fbt2/XCCy/o6aef1oQJE7Rw4UIdPnx41JkjNdj5++KLL3T8+PEh50TjHEfDz372M3V3d+uOO+4IjSXSOZbiv5ZHIxHWsTFGJSUluu6665SXlzfovERZz8PNe7Z4reWYXXZ7uIZ7eW5Jmj17tmbPnh3aLigoUGtrq37605/qhhtuGNExY535TNu3b9fkyZO1fPlya3zBggVasGBBaHvhwoW68sor9atf/Uq//OUvo5I5EuE+v7PHY32OR+rpp5+Wz+fT888/r2nTpoXGE+0cJ8paHolEWMcPPfSQ/v73v+uVV14559xEWM+R5O0Xz7Uct2cMkV6eezALFiywmtLj8Yz6mIMZTWZjjH7/+99r5cqVGj9+/JBzU1JSdPXVV8flf7ODnb/U1FRdeOGFQ86JxjkejWeeeUb33nuv/vjHP2rJkiVDzo3nOR7MWK7lkUqEdfzwww/rhRde0Msvv6zp06cPOTcR1nMkefvFey3HrRiidXnu5uZmZWZmhrYLCgoGHLO+vj4ql/weTeaGhga99957uvfee8/5OMYYvfXWW9bnNVYGO3/z588P/SnVWJ7jkXr66ad19913q7a2Vrfddts558fzHA9mLNfySMVzHRtj9NBDD+nZZ5/VX/7yF+Xk5JzzPvFczyPJKyXIWo7pW9vnsGPHDuN0Os3WrVvNO++8Y4qLi815550X+kmHtWvXmpUrV4bm//znPzd1dXXmX//6l/nHP/5h1q5daySZnTt3hua8+uqrZty4cWbDhg3m3XffNRs2bDCpqalm//79ccnc77vf/a7Jz88Pe0yfz2d2795t/v3vf5vm5mZzzz33mNTUVPP666+POm9XV5dpbm42zc3NRpKpqqoyzc3N5siRI2Hzvv/++yYtLc386Ec/Mu+8847ZunWrcTqd5s9//nNoTqzPcaSZa2trTWpqqnnqqadMW1tb6Hby5MnQnFie45FkjvdajjRvv3itY2OM+f73v28mTZpk9u7da/079/T0hOYk0noeSd5EWMvGGBPXYjDGmKeeesrMmDHDjB8/3lx55ZXWj3KtWrXKLFq0KLS9ceNGM3PmTDNhwgRzwQUXmOuuu87s2rVrwDH/9Kc/mdmzZxun02nmzJljfbGNdWZjjDl58qSZOHGi2bJlS9jjFRcXm4svvtiMHz/eTJ061Xi9XtPY2BiVrP0/Fnn2bdWqVYPm3bt3r5k3b54ZP368ueSSS8zmzZsHHDeW5zjSzIsWLRpyvjGxPccjyRzvtTySdRHPdWyMGfRvGG/bti00J5HW80jyJsJaNsYYLrsNALDE/ZIYAIDEQjEAACwUAwDAQjEAACwUAwDAQjEAACwUAwDAQjEAACwUAwDAQjEAACwUAwDAQjEAACz/D+Xd4mybZCu2AAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 400x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(f\"Year coverage: {conv_rate_year['launch_announced'].sort_values().unique()}\")\n",
    "conversion_rate.hist(figsize=(4, 3))\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 104,
   "metadata": {
    "id": "RJq6-v5OtSeU",
    "outputId": "99fdb20f-2f48-4d0c-e18d-3cc069462374",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lower bound: 0.55\n",
      "Upper bound: 1.56\n",
      "Number of outliers: 8\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>conversion_rate</th>\n",
       "      <th>eur</th>\n",
       "      <th>usd</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>169</th>\n",
       "      <td>1.589414</td>\n",
       "      <td>526.08</td>\n",
       "      <td>330.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1052</th>\n",
       "      <td>1.589414</td>\n",
       "      <td>526.08</td>\n",
       "      <td>330.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1385</th>\n",
       "      <td>1.717929</td>\n",
       "      <td>749.00</td>\n",
       "      <td>435.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>502</th>\n",
       "      <td>1.717929</td>\n",
       "      <td>749.00</td>\n",
       "      <td>435.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>184</th>\n",
       "      <td>1.972696</td>\n",
       "      <td>578.00</td>\n",
       "      <td>293.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1067</th>\n",
       "      <td>1.972696</td>\n",
       "      <td>578.00</td>\n",
       "      <td>293.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1567</th>\n",
       "      <td>2.211872</td>\n",
       "      <td>1172.27</td>\n",
       "      <td>529.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1493</th>\n",
       "      <td>2.286599</td>\n",
       "      <td>997.14</td>\n",
       "      <td>436.08</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      conversion_rate      eur     usd\n",
       "169          1.589414   526.08  330.99\n",
       "1052         1.589414   526.08  330.99\n",
       "1385         1.717929   749.00  435.99\n",
       "502          1.717929   749.00  435.99\n",
       "184          1.972696   578.00  293.00\n",
       "1067         1.972696   578.00  293.00\n",
       "1567         2.211872  1172.27  529.99\n",
       "1493         2.286599   997.14  436.08"
      ]
     },
     "execution_count": 104,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_concat = pd.concat([conversion_rate, prices_ext.loc[conversion_rate.index]], axis=1)\n",
    "util_data.compute_outlier(df_concat, \"conversion_rate\").sort_values(by=\"conversion_rate\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 100 candidates, totalling 500 fits\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-17 {color: black;background-color: white;}#sk-container-id-17 pre{padding: 0;}#sk-container-id-17 div.sk-toggleable {background-color: white;}#sk-container-id-17 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-17 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-17 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-17 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-17 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-17 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-17 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-17 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-17 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-17 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-17 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-17 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-17 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-17 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-17 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-17 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-17 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-17 div.sk-item {position: relative;z-index: 1;}#sk-container-id-17 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-17 div.sk-item::before, #sk-container-id-17 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-17 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-17 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-17 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-17 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-17 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-17 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-17 div.sk-label-container {text-align: center;}#sk-container-id-17 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-17 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-17\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={&#x27;alpha&#x27;: array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        &#x27;epsilon&#x27;: array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        &#x27;fit_intercept&#x27;: [False],\n",
       "                                        &#x27;max_iter&#x27;: [100],\n",
       "                                        &#x27;tol&#x27;: array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit=&#x27;r2&#x27;, scoring=[&#x27;neg_mean_absolute_error&#x27;, &#x27;r2&#x27;],\n",
       "                   verbose=2)</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-49\" type=\"checkbox\" ><label for=\"sk-estimator-id-49\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">RandomizedSearchCV</label><div class=\"sk-toggleable__content\"><pre>RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={&#x27;alpha&#x27;: array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        &#x27;epsilon&#x27;: array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        &#x27;fit_intercept&#x27;: [False],\n",
       "                                        &#x27;max_iter&#x27;: [100],\n",
       "                                        &#x27;tol&#x27;: array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit=&#x27;r2&#x27;, scoring=[&#x27;neg_mean_absolute_error&#x27;, &#x27;r2&#x27;],\n",
       "                   verbose=2)</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-50\" type=\"checkbox\" ><label for=\"sk-estimator-id-50\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">estimator: HuberRegressor</label><div class=\"sk-toggleable__content\"><pre>HuberRegressor()</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-51\" type=\"checkbox\" ><label for=\"sk-estimator-id-51\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">HuberRegressor</label><div class=\"sk-toggleable__content\"><pre>HuberRegressor()</pre></div></div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={'alpha': array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        'epsilon': array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        'fit_intercept': [False],\n",
       "                                        'max_iter': [100],\n",
       "                                        'tol': array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit='r2', scoring=['neg_mean_absolute_error', 'r2'],\n",
       "                   verbose=2)"
      ]
     },
     "execution_count": 159,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# split in train and test\n",
    "# train, test = util_data.split_train_test(prices_ext, test_size=0.2, rng=rng)\n",
    "X_train, y_train = train[curr.lower()].values.reshape(-1, 1), train[\"eur\"].values\n",
    "X_test, y_test = test[curr.lower()].values.reshape(-1, 1), test[\"eur\"].values\n",
    "\n",
    "params = {\n",
    "    \"epsilon\": np.linspace(1, 2, num=10),\n",
    "    \"max_iter\": [100],\n",
    "    \"alpha\": np.linspace(0.0001, 0.1, num=20),\n",
    "    \"fit_intercept\": [False],\n",
    "    \"tol\": np.linspace(1e-5, 1e-4, num=10)\n",
    "}\n",
    "scoring = [\"neg_mean_absolute_error\", \"r2\"]\n",
    "rscv = RandomizedSearchCV(HuberRegressor(), params, n_iter=100, scoring=scoring, n_jobs=-1,\n",
    "                          refit=\"r2\", cv=5, verbose=2, error_score=np.nan)\n",
    "rscv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "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>std_test_r2</th>\n",
       "      <th>mean_test_r2</th>\n",
       "      <th>mean_test_neg_mean_absolute_error</th>\n",
       "      <th>std_test_neg_mean_absolute_error</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>87</th>\n",
       "      <td>0.194146</td>\n",
       "      <td>0.681986</td>\n",
       "      <td>-86.483241</td>\n",
       "      <td>25.509428</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.194146</td>\n",
       "      <td>0.681986</td>\n",
       "      <td>-86.483253</td>\n",
       "      <td>25.509419</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>53</th>\n",
       "      <td>0.194146</td>\n",
       "      <td>0.681986</td>\n",
       "      <td>-86.483253</td>\n",
       "      <td>25.509419</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>0.194146</td>\n",
       "      <td>0.681986</td>\n",
       "      <td>-86.483276</td>\n",
       "      <td>25.509399</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>58</th>\n",
       "      <td>0.194146</td>\n",
       "      <td>0.681986</td>\n",
       "      <td>-86.483276</td>\n",
       "      <td>25.509399</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    std_test_r2  mean_test_r2  mean_test_neg_mean_absolute_error  \\\n",
       "87     0.194146      0.681986                         -86.483241   \n",
       "15     0.194146      0.681986                         -86.483253   \n",
       "53     0.194146      0.681986                         -86.483253   \n",
       "16     0.194146      0.681986                         -86.483276   \n",
       "58     0.194146      0.681986                         -86.483276   \n",
       "\n",
       "    std_test_neg_mean_absolute_error  \n",
       "87                         25.509428  \n",
       "15                         25.509419  \n",
       "53                         25.509419  \n",
       "16                         25.509399  \n",
       "58                         25.509399  "
      ]
     },
     "execution_count": 166,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "usd_results = pd.DataFrame(rscv.cv_results_).sort_values(\"rank_test_neg_mean_absolute_error\")\n",
    "usd_sorted_r2= usd_results.sort_values(by=\"std_test_r2\")\n",
    "usd_sorted_rmse = usd_results.sort_values(by=\"std_test_neg_mean_absolute_error\")\n",
    "cols = [\"std_test_r2\", \"mean_test_r2\", \"mean_test_neg_mean_absolute_error\", \"std_test_neg_mean_absolute_error\"]\n",
    "usd_results.iloc[:5][cols]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "def print_best_n_search(df: pd.DataFrame, n: int, metrics: dict):\n",
    "    for i in range(n):\n",
    "        print(f\"BEST {i+1} SEARCH:\")\n",
    "        for name, metric in metrics.items():\n",
    "            print(f\"STD {name}: {df.iloc[i][f'std_test_{metric}']}\")\n",
    "            print(f\"MEAN {name}: {df.iloc[i][f'mean_test_{metric}']}\")\n",
    "            print(f\"Rank {name}: {df.iloc[i][f'rank_test_{metric}']}\", end=\"\\n\\n\")\n",
    "        print(\"---\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 167,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---BEST R2---\n",
      "BEST 1 SEARCH:\n",
      "STD R2: 0.1778767023446916\n",
      "MEAN R2: 0.6924911185895026\n",
      "Rank R2: 1\n",
      "\n",
      "STD MAE: 25.571544872898752\n",
      "MEAN MAE: -87.05065615293589\n",
      "Rank MAE: 100\n",
      "\n",
      "---\n",
      "BEST 2 SEARCH:\n",
      "STD R2: 0.17787671308010156\n",
      "MEAN R2: 0.6924911146526791\n",
      "Rank R2: 2\n",
      "\n",
      "STD MAE: 25.57154503719598\n",
      "MEAN MAE: -87.05065477317567\n",
      "Rank MAE: 99\n",
      "\n",
      "---\n",
      "BEST 3 SEARCH:\n",
      "STD R2: 0.1778767362508994\n",
      "MEAN R2: 0.692491097064998\n",
      "Rank R2: 3\n",
      "\n",
      "STD MAE: 25.571545056966254\n",
      "MEAN MAE: -87.05065381259355\n",
      "Rank MAE: 98\n",
      "\n",
      "---\n",
      "---BEST RMSE---\n",
      "BEST 1 SEARCH:\n",
      "STD R2: 0.17908882646446458\n",
      "MEAN R2: 0.6883626019911133\n",
      "Rank R2: 87\n",
      "\n",
      "STD MAE: 24.99210275598061\n",
      "MEAN MAE: -86.69142956165892\n",
      "Rank MAE: 71\n",
      "\n",
      "---\n",
      "BEST 2 SEARCH:\n",
      "STD R2: 0.1790887444312414\n",
      "MEAN R2: 0.6883629400261851\n",
      "Rank R2: 85\n",
      "\n",
      "STD MAE: 24.992151307179476\n",
      "MEAN MAE: -86.6913682323731\n",
      "Rank MAE: 70\n",
      "\n",
      "---\n",
      "BEST 3 SEARCH:\n",
      "STD R2: 0.1790886046921912\n",
      "MEAN R2: 0.6883632294935613\n",
      "Rank R2: 82\n",
      "\n",
      "STD MAE: 24.992172539469667\n",
      "MEAN MAE: -86.69133516300691\n",
      "Rank MAE: 69\n",
      "\n",
      "---\n"
     ]
    }
   ],
   "source": [
    "metrics = {\"R2\": \"r2\", \"MAE\": \"neg_mean_absolute_error\"}\n",
    "print(f\"---BEST R2---\")\n",
    "print_best_n_search(usd_sorted_r2, 3, metrics)\n",
    "print(f\"---BEST MAE---\")\n",
    "print_best_n_search(usd_sorted_rmse, 3, metrics)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The results shows that the conversion rate of the Huber regressor is approximately *1.058*, similar to the conversion rate retrieved from the European Central Bank, thus we decided to use the latter."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 177,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "rate_train = train[\"eur\"] / train[curr.lower()]\n",
    "\n",
    "model_median = util_models.FixedLinearRegressor(\n",
    "    rate_train.median(), fit_intercept=False\n",
    ")\n",
    "model_fit = HuberRegressor(**usd_results.iloc[0][\"params\"])\n",
    "model_rate = util_models.FixedLinearRegressor(0.8694*1.22, fit_intercept=False)\n",
    "\n",
    "models = [model_median, model_fit, model_rate]\n",
    "names = [\"median\", \"linear regressor\", \"online_rate\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "metadata": {
    "id": "Az0kdpditSeU",
    "outputId": "2fc3ab7f-100a-46b9-f6bf-65f44d87a6ad",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "median coefficients: [1.07572383]\n",
      "median RMSE: 97.70\n",
      "median r2-score: 0.85\n",
      "median MAE: 58.81\n",
      "\n",
      "linear regressor coefficients: [1.05882697]\n",
      "linear regressor RMSE: 95.16\n",
      "linear regressor r2-score: 0.86\n",
      "linear regressor MAE: 57.22\n",
      "\n",
      "online_rate coefficients: [1.060668]\n",
      "online_rate RMSE: 95.42\n",
      "online_rate r2-score: 0.86\n",
      "online_rate MAE: 57.40\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_preds = util_models.compare_models(X_train, y_train, X_test, y_test, models, names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the data\n",
    "util_plot.plot_compare_models(\n",
    "    X_test, y_test, y_preds, names, f\"{curr}/EUR conversion rates\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SQVUY8M4tSeU",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Conversion rate between GBP and EUR"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The conversion rate retrieved from the ECB *1.1315*, while the median conversion rate is *1.0792*.\n",
    "\n",
    "To estimate the conversion rate between GBP and EUR we still perform regression with the Huber regressor model (outliers are still present)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 193,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of prices: 187\n"
     ]
    }
   ],
   "source": [
    "curr = \"GBP\"\n",
    "prices_ext, conversion_rate = util_preprocess.extract_prices_rate(curr, misc_price, first_eur=False)\n",
    "\n",
    "# add the year to the conversion rate\n",
    "conv_rate_year = pd.concat([conversion_rate, ser_year.loc[conversion_rate.index]], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Year coverage: [2018. 2019. 2020.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(f\"Year coverage: {conv_rate_year['launch_announced'].sort_values().unique()}\")\n",
    "conversion_rate.hist(figsize=(4, 3))\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lower bound: 0.76\n",
      "Upper bound: 1.39\n",
      "Number of outliers: 16\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>conversion_rate</th>\n",
       "      <th>gbp</th>\n",
       "      <th>eur</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>184</th>\n",
       "      <td>0.680000</td>\n",
       "      <td>850.00</td>\n",
       "      <td>578.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1067</th>\n",
       "      <td>0.680000</td>\n",
       "      <td>850.00</td>\n",
       "      <td>578.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1500</th>\n",
       "      <td>0.759422</td>\n",
       "      <td>283.11</td>\n",
       "      <td>215.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>962</th>\n",
       "      <td>1.408325</td>\n",
       "      <td>528.99</td>\n",
       "      <td>744.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>79</th>\n",
       "      <td>1.408325</td>\n",
       "      <td>528.99</td>\n",
       "      <td>744.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1502</th>\n",
       "      <td>1.408451</td>\n",
       "      <td>119.99</td>\n",
       "      <td>169.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>847</th>\n",
       "      <td>1.428587</td>\n",
       "      <td>279.99</td>\n",
       "      <td>399.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>820</th>\n",
       "      <td>1.438857</td>\n",
       "      <td>555.99</td>\n",
       "      <td>799.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>511</th>\n",
       "      <td>1.453978</td>\n",
       "      <td>204.90</td>\n",
       "      <td>297.92</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1394</th>\n",
       "      <td>1.453978</td>\n",
       "      <td>204.90</td>\n",
       "      <td>297.92</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1191</th>\n",
       "      <td>1.640974</td>\n",
       "      <td>195.00</td>\n",
       "      <td>319.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>308</th>\n",
       "      <td>1.640974</td>\n",
       "      <td>195.00</td>\n",
       "      <td>319.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>389</th>\n",
       "      <td>1.646507</td>\n",
       "      <td>64.98</td>\n",
       "      <td>106.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1272</th>\n",
       "      <td>1.646507</td>\n",
       "      <td>64.98</td>\n",
       "      <td>106.99</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>298</th>\n",
       "      <td>36.537102</td>\n",
       "      <td>42.99</td>\n",
       "      <td>1570.73</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1181</th>\n",
       "      <td>36.537102</td>\n",
       "      <td>42.99</td>\n",
       "      <td>1570.73</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      conversion_rate     gbp      eur\n",
       "184          0.680000  850.00   578.00\n",
       "1067         0.680000  850.00   578.00\n",
       "1500         0.759422  283.11   215.00\n",
       "962          1.408325  528.99   744.99\n",
       "79           1.408325  528.99   744.99\n",
       "1502         1.408451  119.99   169.00\n",
       "847          1.428587  279.99   399.99\n",
       "820          1.438857  555.99   799.99\n",
       "511          1.453978  204.90   297.92\n",
       "1394         1.453978  204.90   297.92\n",
       "1191         1.640974  195.00   319.99\n",
       "308          1.640974  195.00   319.99\n",
       "389          1.646507   64.98   106.99\n",
       "1272         1.646507   64.98   106.99\n",
       "298         36.537102   42.99  1570.73\n",
       "1181        36.537102   42.99  1570.73"
      ]
     },
     "execution_count": 200,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_concat = pd.concat([conversion_rate, prices_ext.loc[conversion_rate.index]], axis=1)\n",
    "util_data.compute_outlier(df_concat, \"conversion_rate\").sort_values(by=\"conversion_rate\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "metadata": {
    "id": "7SvXXI3KtSeU",
    "outputId": "84344033-10e8-492d-d725-59fa5dec96eb",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "conversion_rate.plot.box(whis=1.5)\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 222,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Fitting 5 folds for each of 100 candidates, totalling 500 fits\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<style>#sk-container-id-27 {color: black;background-color: white;}#sk-container-id-27 pre{padding: 0;}#sk-container-id-27 div.sk-toggleable {background-color: white;}#sk-container-id-27 label.sk-toggleable__label {cursor: pointer;display: block;width: 100%;margin-bottom: 0;padding: 0.3em;box-sizing: border-box;text-align: center;}#sk-container-id-27 label.sk-toggleable__label-arrow:before {content: \"▸\";float: left;margin-right: 0.25em;color: #696969;}#sk-container-id-27 label.sk-toggleable__label-arrow:hover:before {color: black;}#sk-container-id-27 div.sk-estimator:hover label.sk-toggleable__label-arrow:before {color: black;}#sk-container-id-27 div.sk-toggleable__content {max-height: 0;max-width: 0;overflow: hidden;text-align: left;background-color: #f0f8ff;}#sk-container-id-27 div.sk-toggleable__content pre {margin: 0.2em;color: black;border-radius: 0.25em;background-color: #f0f8ff;}#sk-container-id-27 input.sk-toggleable__control:checked~div.sk-toggleable__content {max-height: 200px;max-width: 100%;overflow: auto;}#sk-container-id-27 input.sk-toggleable__control:checked~label.sk-toggleable__label-arrow:before {content: \"▾\";}#sk-container-id-27 div.sk-estimator input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-27 div.sk-label input.sk-toggleable__control:checked~label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-27 input.sk-hidden--visually {border: 0;clip: rect(1px 1px 1px 1px);clip: rect(1px, 1px, 1px, 1px);height: 1px;margin: -1px;overflow: hidden;padding: 0;position: absolute;width: 1px;}#sk-container-id-27 div.sk-estimator {font-family: monospace;background-color: #f0f8ff;border: 1px dotted black;border-radius: 0.25em;box-sizing: border-box;margin-bottom: 0.5em;}#sk-container-id-27 div.sk-estimator:hover {background-color: #d4ebff;}#sk-container-id-27 div.sk-parallel-item::after {content: \"\";width: 100%;border-bottom: 1px solid gray;flex-grow: 1;}#sk-container-id-27 div.sk-label:hover label.sk-toggleable__label {background-color: #d4ebff;}#sk-container-id-27 div.sk-serial::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: 0;}#sk-container-id-27 div.sk-serial {display: flex;flex-direction: column;align-items: center;background-color: white;padding-right: 0.2em;padding-left: 0.2em;position: relative;}#sk-container-id-27 div.sk-item {position: relative;z-index: 1;}#sk-container-id-27 div.sk-parallel {display: flex;align-items: stretch;justify-content: center;background-color: white;position: relative;}#sk-container-id-27 div.sk-item::before, #sk-container-id-27 div.sk-parallel-item::before {content: \"\";position: absolute;border-left: 1px solid gray;box-sizing: border-box;top: 0;bottom: 0;left: 50%;z-index: -1;}#sk-container-id-27 div.sk-parallel-item {display: flex;flex-direction: column;z-index: 1;position: relative;background-color: white;}#sk-container-id-27 div.sk-parallel-item:first-child::after {align-self: flex-end;width: 50%;}#sk-container-id-27 div.sk-parallel-item:last-child::after {align-self: flex-start;width: 50%;}#sk-container-id-27 div.sk-parallel-item:only-child::after {width: 0;}#sk-container-id-27 div.sk-dashed-wrapped {border: 1px dashed gray;margin: 0 0.4em 0.5em 0.4em;box-sizing: border-box;padding-bottom: 0.4em;background-color: white;}#sk-container-id-27 div.sk-label label {font-family: monospace;font-weight: bold;display: inline-block;line-height: 1.2em;}#sk-container-id-27 div.sk-label-container {text-align: center;}#sk-container-id-27 div.sk-container {/* jupyter's `normalize.less` sets `[hidden] { display: none; }` but bootstrap.min.css set `[hidden] { display: none !important; }` so we also need the `!important` here to be able to override the default hidden behavior on the sphinx rendered scikit-learn.org. See: https://github.com/scikit-learn/scikit-learn/issues/21755 */display: inline-block !important;position: relative;}#sk-container-id-27 div.sk-text-repr-fallback {display: none;}</style><div id=\"sk-container-id-27\" class=\"sk-top-container\"><div class=\"sk-text-repr-fallback\"><pre>RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={&#x27;alpha&#x27;: array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        &#x27;epsilon&#x27;: array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        &#x27;fit_intercept&#x27;: [False],\n",
       "                                        &#x27;max_iter&#x27;: [100],\n",
       "                                        &#x27;tol&#x27;: array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit=&#x27;r2&#x27;, scoring=[&#x27;neg_mean_absolute_error&#x27;, &#x27;r2&#x27;],\n",
       "                   verbose=2)</pre><b>In a Jupyter environment, please rerun this cell to show the HTML representation or trust the notebook. <br />On GitHub, the HTML representation is unable to render, please try loading this page with nbviewer.org.</b></div><div class=\"sk-container\" hidden><div class=\"sk-item sk-dashed-wrapped\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-79\" type=\"checkbox\" ><label for=\"sk-estimator-id-79\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">RandomizedSearchCV</label><div class=\"sk-toggleable__content\"><pre>RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={&#x27;alpha&#x27;: array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        &#x27;epsilon&#x27;: array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        &#x27;fit_intercept&#x27;: [False],\n",
       "                                        &#x27;max_iter&#x27;: [100],\n",
       "                                        &#x27;tol&#x27;: array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit=&#x27;r2&#x27;, scoring=[&#x27;neg_mean_absolute_error&#x27;, &#x27;r2&#x27;],\n",
       "                   verbose=2)</pre></div></div></div><div class=\"sk-parallel\"><div class=\"sk-parallel-item\"><div class=\"sk-item\"><div class=\"sk-label-container\"><div class=\"sk-label sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-80\" type=\"checkbox\" ><label for=\"sk-estimator-id-80\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">estimator: HuberRegressor</label><div class=\"sk-toggleable__content\"><pre>HuberRegressor()</pre></div></div></div><div class=\"sk-serial\"><div class=\"sk-item\"><div class=\"sk-estimator sk-toggleable\"><input class=\"sk-toggleable__control sk-hidden--visually\" id=\"sk-estimator-id-81\" type=\"checkbox\" ><label for=\"sk-estimator-id-81\" class=\"sk-toggleable__label sk-toggleable__label-arrow\">HuberRegressor</label><div class=\"sk-toggleable__content\"><pre>HuberRegressor()</pre></div></div></div></div></div></div></div></div></div></div>"
      ],
      "text/plain": [
       "RandomizedSearchCV(cv=5, estimator=HuberRegressor(), n_iter=100, n_jobs=-1,\n",
       "                   param_distributions={'alpha': array([0.0001    , 0.00535789, 0.01061579, 0.01587368, 0.02113158,\n",
       "       0.02638947, 0.03164737, 0.03690526, 0.04216316, 0.04742105,\n",
       "       0.05267895, 0.05793684, 0.06319474, 0.06845263, 0.07371053,\n",
       "       0.07896842, 0.08422632, 0.08948421, 0.09474211, 0.1       ]),\n",
       "                                        'epsilon': array([1.        , 1.11111111, 1.22222222, 1.33333333, 1.44444444,\n",
       "       1.55555556, 1.66666667, 1.77777778, 1.88888889, 2.        ]),\n",
       "                                        'fit_intercept': [False],\n",
       "                                        'max_iter': [100],\n",
       "                                        'tol': array([1.e-05, 2.e-05, 3.e-05, 4.e-05, 5.e-05, 6.e-05, 7.e-05, 8.e-05,\n",
       "       9.e-05, 1.e-04])},\n",
       "                   refit='r2', scoring=['neg_mean_absolute_error', 'r2'],\n",
       "                   verbose=2)"
      ]
     },
     "execution_count": 222,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# split in train and test\n",
    "train, test = util_data.split_train_test(prices_ext, test_size=0.2, rng=rng)\n",
    "X_train, y_train = train[curr.lower()].values.reshape(-1, 1), train[\"eur\"].values\n",
    "X_test, y_test = test[curr.lower()].values.reshape(-1, 1), test[\"eur\"].values\n",
    "\n",
    "params = {\n",
    "    \"epsilon\": np.linspace(1, 2, num=10),\n",
    "    \"max_iter\": [100],\n",
    "    \"alpha\": np.linspace(0.0001, 0.1, num=20),\n",
    "    \"fit_intercept\": [False],\n",
    "    \"tol\": np.linspace(1e-5, 1e-4, num=10)\n",
    "}\n",
    "scoring = [\"neg_mean_absolute_error\", \"r2\"]\n",
    "rscv = RandomizedSearchCV(HuberRegressor(), params, n_iter=100, scoring=scoring, n_jobs=-1,\n",
    "                          refit=\"r2\", cv=5, verbose=2, error_score=np.nan)\n",
    "rscv.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 211,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "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>std_test_r2</th>\n",
       "      <th>mean_test_r2</th>\n",
       "      <th>mean_test_neg_mean_absolute_error</th>\n",
       "      <th>std_test_neg_mean_absolute_error</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>0.440197</td>\n",
       "      <td>0.718153</td>\n",
       "      <td>-58.003204</td>\n",
       "      <td>20.585625</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.440197</td>\n",
       "      <td>0.718153</td>\n",
       "      <td>-58.003205</td>\n",
       "      <td>20.585623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>0.440197</td>\n",
       "      <td>0.718153</td>\n",
       "      <td>-58.003205</td>\n",
       "      <td>20.585623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>74</th>\n",
       "      <td>0.440197</td>\n",
       "      <td>0.718153</td>\n",
       "      <td>-58.003210</td>\n",
       "      <td>20.585624</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>87</th>\n",
       "      <td>0.440197</td>\n",
       "      <td>0.718153</td>\n",
       "      <td>-58.003211</td>\n",
       "      <td>20.585605</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    std_test_r2  mean_test_r2  mean_test_neg_mean_absolute_error  \\\n",
       "22     0.440197      0.718153                         -58.003204   \n",
       "7      0.440197      0.718153                         -58.003205   \n",
       "15     0.440197      0.718153                         -58.003205   \n",
       "74     0.440197      0.718153                         -58.003210   \n",
       "87     0.440197      0.718153                         -58.003211   \n",
       "\n",
       "    std_test_neg_mean_absolute_error  \n",
       "22                         20.585625  \n",
       "7                          20.585623  \n",
       "15                         20.585623  \n",
       "74                         20.585624  \n",
       "87                         20.585605  "
      ]
     },
     "execution_count": 211,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gbp_results = pd.DataFrame(rscv.cv_results_).sort_values(by=\"rank_test_neg_mean_absolute_error\")\n",
    "gbp_sorted_r2= gbp_results.sort_values(by=\"std_test_r2\")\n",
    "gbp_sorted_rmse = gbp_results.sort_values(by=\"std_test_neg_mean_absolute_error\")\n",
    "gbp_results.iloc[:5][cols]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 225,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---BEST R2---\n",
      "BEST 1 SEARCH:\n",
      "STD R2: 0.439241409264153\n",
      "MEAN R2: 0.7178813901401159\n",
      "Rank R2: 61\n",
      "\n",
      "STD MAE: 20.326537064320508\n",
      "MEAN MAE: -58.369182725078545\n",
      "Rank MAE: 74\n",
      "\n",
      "---\n",
      "BEST 2 SEARCH:\n",
      "STD R2: 0.43924141284589513\n",
      "MEAN R2: 0.7178813650013284\n",
      "Rank R2: 62\n",
      "\n",
      "STD MAE: 20.32654584400644\n",
      "MEAN MAE: -58.36918777903164\n",
      "Rank MAE: 75\n",
      "\n",
      "---\n",
      "BEST 3 SEARCH:\n",
      "STD R2: 0.43924141437120595\n",
      "MEAN R2: 0.7178813542350674\n",
      "Rank R2: 63\n",
      "\n",
      "STD MAE: 20.326549608846896\n",
      "MEAN MAE: -58.369189944910886\n",
      "Rank MAE: 76\n",
      "\n",
      "---\n",
      "---BEST MAE---\n",
      "BEST 1 SEARCH:\n",
      "STD R2: 0.4398075685382518\n",
      "MEAN R2: 0.7182998386397993\n",
      "Rank R2: 27\n",
      "\n",
      "STD MAE: 20.09657490084541\n",
      "MEAN MAE: -58.15045998673715\n",
      "Rank MAE: 47\n",
      "\n",
      "---\n",
      "BEST 2 SEARCH:\n",
      "STD R2: 0.4398075685382518\n",
      "MEAN R2: 0.7182998386397993\n",
      "Rank R2: 27\n",
      "\n",
      "STD MAE: 20.09657490084541\n",
      "MEAN MAE: -58.15045998673715\n",
      "Rank MAE: 47\n",
      "\n",
      "---\n",
      "BEST 3 SEARCH:\n",
      "STD R2: 0.4398075698624437\n",
      "MEAN R2: 0.7182998373527785\n",
      "Rank R2: 29\n",
      "\n",
      "STD MAE: 20.09657585705557\n",
      "MEAN MAE: -58.15046002677116\n",
      "Rank MAE: 49\n",
      "\n",
      "---\n"
     ]
    }
   ],
   "source": [
    "print(f\"---BEST R2---\")\n",
    "print_best_n_search(gbp_sorted_r2, 3, metrics)\n",
    "print(f\"---BEST MAE---\")\n",
    "print_best_n_search(gbp_sorted_rmse, 3, metrics)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The results shows that the conversion rate of the Huber regressor is *1.0458*, which is the similar as the median conversion rate, however we chose the former."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 226,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "rate_train = train[\"eur\"] / train[curr.lower()]\n",
    "\n",
    "model_median = util_models.FixedLinearRegressor(\n",
    "    rate_train.median(), fit_intercept=False\n",
    ")\n",
    "model_fit = HuberRegressor(**gbp_results.iloc[0][\"params\"])\n",
    "model_rate = util_models.FixedLinearRegressor(1.1315, fit_intercept=False)\n",
    "\n",
    "models = [model_median, model_fit, model_rate]\n",
    "names = [\"median\", \"linear regressor\", \"online_rate\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 227,
   "metadata": {
    "id": "UQ-reTBStSeV",
    "outputId": "09f67dea-e25b-44df-90bc-a90fcf3536ae",
    "scrolled": false,
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "median coefficients: [1.07927112]\n",
      "median RMSE: 254.09\n",
      "median r2-score: 0.40\n",
      "median MAE: 78.58\n",
      "\n",
      "linear regressor coefficients: [1.04589785]\n",
      "linear regressor RMSE: 254.94\n",
      "linear regressor r2-score: 0.40\n",
      "linear regressor MAE: 78.40\n",
      "\n",
      "online_rate coefficients: [1.1315]\n",
      "online_rate RMSE: 254.48\n",
      "online_rate r2-score: 0.40\n",
      "online_rate MAE: 82.58\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_preds = util_models.compare_models(X_train, y_train, X_test, y_test, models, names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 228,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the data\n",
    "util_plot.plot_compare_models(\n",
    "    X_test, y_test, y_preds, names, f\"{curr}/EUR conversion rates\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3VSdqB1mtSeQ",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Conversion rate between INR and EUR"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The conversion rate retrieved from the ECB *0.01268*, while the median conversion rate is *0.0114*.\n",
    "\n",
    "To estimate the conversion rate between INR and EUR we perform linear regression (outliers aren't present)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 233,
   "metadata": {
    "id": "1UIFbnx6tSeR",
    "outputId": "567fabaf-7509-46c4-8470-225762ae1e87",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of prices: 40\n"
     ]
    }
   ],
   "source": [
    "curr = \"INR\"\n",
    "prices_ext, conversion_rate = util_preprocess.extract_prices_rate(curr, misc_price)\n",
    "\n",
    "# add the year to the conversion rate\n",
    "conv_rate_year = pd.concat([conversion_rate, ser_year.loc[conversion_rate.index]], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 234,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Year coverage: [2019. 2020.]\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 400x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(f\"Year coverage: {conv_rate_year['launch_announced'].sort_values().unique()}\")\n",
    "conversion_rate.hist(figsize=(4, 3))\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 235,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Lower bound: 0.01\n",
      "Upper bound: 0.02\n",
      "Number of outliers: 0\n"
     ]
    },
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>conversion_rate</th>\n",
       "      <th>eur</th>\n",
       "      <th>inr</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: [conversion_rate, eur, inr]\n",
       "Index: []"
      ]
     },
     "execution_count": 235,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_concat = pd.concat([conversion_rate, prices_ext.loc[conversion_rate.index]], axis=1)\n",
    "util_data.compute_outlier(df_concat, \"conversion_rate\").sort_values(by=\"conversion_rate\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 236,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAnYAAAHWCAYAAAD6oMSKAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAy+0lEQVR4nO3dfVhVdb7//9eWu20iHMMEUeQmp8LQ0o1fB42hmQoHqyvnMB0qc+x4c4ajkyE1JXqcuuwGuyIPOSocjDTHSZ1zcZqsKKUmyRKbAaFMuUqNgrg5DFbgTXHn+v3Rz31mt7fIJg39+Hxc17qG9VnvtT7vxVwzvfps1to2y7IsAQAA4II3oL8bAAAAwNlBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQ/j2dwM/pJMnT6qhoUGDBw+WzWbr73YAAADOyLIsHT16VOHh4RowoOc1uYsq2DU0NCgiIqK/2wAAAPBaXV2dRo4c2WPNRRXsBg8eLOnbX0xQUFA/dwMAAHBmbW1tioiIcOaYnlxUwe7Ux69BQUEEOwAAcEHpzZ+R8fAEAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiiT8Fu7dq1io6Olt1ul8Ph0K5du3qsLy0tlcPhkN1uV0xMjPLz812O79+/X6mpqYqKipLNZlNubq7bNU4d++62YMGCvtwCAACAcbwOdlu3blVGRoaWLl2qyspKJSYmKiUlRbW1tR7ra2pqNG3aNCUmJqqyslJLlizRwoULVVRU5Kw5ceKEYmJitGLFCoWFhXm8zt/+9jc1NjY6t5KSEknS7bff7u0tAAAAGMlmWZblzQmTJk3ShAkTlJeX5xyLjY3V9OnTlZ2d7Vb/0EMPadu2baqurnaOpaen6/3331dZWZlbfVRUlDIyMpSRkdFjHxkZGXrllVd08ODBXr2JWfr2KzmCg4PV2trKN08AAIALgjf5xasVu46ODlVUVCg5OdllPDk5Wbt37/Z4TllZmVv91KlTVV5ers7OTm+md+lj06ZNmj17dq9DHQAAgOm8+q7YlpYWdXd3KzQ01GU8NDRUTU1NHs9pamryWN/V1aWWlhYNHz7cy5alP//5z/rqq690zz339FjX3t6u9vZ2535bW5vXcwEAAFwo+vTwxHdXySzL6nHlzFO9p/HeKiwsVEpKisLDw3usy87OVnBwsHOLiIjo03wAAAAXAq+C3dChQ+Xj4+O2Otfc3Oy2KndKWFiYx3pfX1+FhIR42a702Wef6Y033tDcuXPPWJuVlaXW1lbnVldX5/V8AAAAFwqvgp2/v78cDofzidRTSkpKNHnyZI/nJCQkuNXv2LFD8fHx8vPz87Jdaf369Ro2bJhuvvnmM9YGBAQoKCjIZQMAADCVV39jJ0mZmZmaOXOm4uPjlZCQoIKCAtXW1io9PV3St6tk9fX12rhxo6Rvn4BdvXq1MjMzNW/ePJWVlamwsFCbN292XrOjo0MHDhxw/lxfX6+qqioFBgZq9OjRzrqTJ09q/fr1mjVrlnx9vW4dACRJX3d06/Dfj53Va37T2a3Pv/xaI4cMlN3P56xe+/LLAjXQ/+xeE4CZvE5HaWlpOnLkiJYvX67GxkbFxcWpuLhYkZGRkqTGxkaXd9pFR0eruLhYixYt0po1axQeHq5Vq1YpNTXVWdPQ0KDx48c793NycpSTk6OkpCTt3LnTOf7GG2+otrZWs2fP7su9AoAk6fDfj+mW37/T32302iv3Xqe4EcH93QaAC4DX77G7kPEeOwDSuVmxO9R8TBlbq5Sbdq1GDws8q9dmxQ64uHmTX/g8E8BFZ6C/zzlbARs9LJDVNQD9pk+vOwEAAMD5h2AHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgiD4Fu7Vr1yo6Olp2u10Oh0O7du3qsb60tFQOh0N2u10xMTHKz893Ob5//36lpqYqKipKNptNubm5Hq9TX1+vu+++WyEhIbrkkkt07bXXqqKioi+3AAAAYByvg93WrVuVkZGhpUuXqrKyUomJiUpJSVFtba3H+pqaGk2bNk2JiYmqrKzUkiVLtHDhQhUVFTlrTpw4oZiYGK1YsUJhYWEer/Pll19qypQp8vPz02uvvaYDBw7o6aef1j/90z95ewsAAABGslmWZXlzwqRJkzRhwgTl5eU5x2JjYzV9+nRlZ2e71T/00EPatm2bqqurnWPp6el6//33VVZW5lYfFRWljIwMZWRkuIwvXrxY77777hlXB3vS1tam4OBgtba2KigoqM/XAYDv+rC+Vbf8/h29cu91ihsR3N/tADCIN/nFqxW7jo4OVVRUKDk52WU8OTlZu3fv9nhOWVmZW/3UqVNVXl6uzs7OXs+9bds2xcfH6/bbb9ewYcM0fvx4rVu3rsdz2tvb1dbW5rIBAACYyqtg19LSou7uboWGhrqMh4aGqqmpyeM5TU1NHuu7urrU0tLS67k/+eQT5eXl6Uc/+pG2b9+u9PR0LVy4UBs3bjztOdnZ2QoODnZuERERvZ4PAADgQtOnhydsNpvLvmVZbmNnqvc03pOTJ09qwoQJeuKJJzR+/Hj9+te/1rx581w+Ev6urKwstba2Ore6urpezwcAAHCh8SrYDR06VD4+Pm6rc83NzW6rcqeEhYV5rPf19VVISEiv5x4+fLjGjBnjMhYbG3vahzYkKSAgQEFBQS4bAACAqbwKdv7+/nI4HCopKXEZLykp0eTJkz2ek5CQ4Fa/Y8cOxcfHy8/Pr9dzT5kyRR999JHL2Mcff6zIyMheXwMAAMBkXn8Um5mZqWeffVbPPfecqqurtWjRItXW1io9PV3Stx9//upXv3LWp6en67PPPlNmZqaqq6v13HPPqbCwUA888ICzpqOjQ1VVVaqqqlJHR4fq6+tVVVWlQ4cOOWsWLVqkPXv26IknntChQ4f0wgsvqKCgQAsWLPg+9w8AAGAOqw/WrFljRUZGWv7+/taECROs0tJS57FZs2ZZSUlJLvU7d+60xo8fb/n7+1tRUVFWXl6ey/GamhpLktv23eu8/PLLVlxcnBUQEGBdddVVVkFBgVd9t7a2WpKs1tZWr84DgDPZ9/lXVuRDr1j7Pv+qv1sBYBhv8ovX77G7kPEeOwDnCu+xA3CunLP32AEAAOD8RbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwRJ+C3dq1axUdHS273S6Hw6Fdu3b1WF9aWiqHwyG73a6YmBjl5+e7HN+/f79SU1MVFRUlm82m3Nxct2s88sgjstlsLltYWFhf2gcAADCS18Fu69atysjI0NKlS1VZWanExESlpKSotrbWY31NTY2mTZumxMREVVZWasmSJVq4cKGKioqcNSdOnFBMTIxWrFjRY1i7+uqr1djY6Nz27dvnbfsAAADG8vX2hJUrV2rOnDmaO3euJCk3N1fbt29XXl6esrOz3erz8/M1atQo5ypcbGysysvLlZOTo9TUVEnSxIkTNXHiREnS4sWLT9+sry+rdAAAAKfh1YpdR0eHKioqlJyc7DKenJys3bt3ezynrKzMrX7q1KkqLy9XZ2enV80ePHhQ4eHhio6O1h133KFPPvnEq/MBAABM5lWwa2lpUXd3t0JDQ13GQ0ND1dTU5PGcpqYmj/VdXV1qaWnp9dyTJk3Sxo0btX37dq1bt05NTU2aPHmyjhw5ctpz2tvb1dbW5rIBAACYqk8PT9hsNpd9y7Lcxs5U72m8JykpKUpNTdXYsWN144036tVXX5UkPf/886c9Jzs7W8HBwc4tIiKi1/MBAABcaLwKdkOHDpWPj4/b6lxzc7PbqtwpYWFhHut9fX0VEhLiZbv/Z9CgQRo7dqwOHjx42pqsrCy1trY6t7q6uj7PBwAAcL7zKtj5+/vL4XCopKTEZbykpESTJ0/2eE5CQoJb/Y4dOxQfHy8/Pz8v2/0/7e3tqq6u1vDhw09bExAQoKCgIJcNAADAVF5/FJuZmalnn31Wzz33nKqrq7Vo0SLV1tYqPT1d0rerZL/61a+c9enp6frss8+UmZmp6upqPffccyosLNQDDzzgrOno6FBVVZWqqqrU0dGh+vp6VVVV6dChQ86aBx54QKWlpaqpqdF7772nX/7yl2pra9OsWbO+z/0DAAAYw+vXnaSlpenIkSNavny5GhsbFRcXp+LiYkVGRkqSGhsbXd5pFx0dreLiYi1atEhr1qxReHi4Vq1a5XzViSQ1NDRo/Pjxzv2cnBzl5OQoKSlJO3fulCR9/vnnuvPOO9XS0qLLLrtMP/7xj7Vnzx7nvAAAABc7m3XqSYaLQFtbm4KDg9Xa2srHsgDOqg/rW3XL79/RK/dep7gRwf3dDgCDeJNf+K5YAAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMESfgt3atWsVHR0tu90uh8OhXbt29VhfWloqh8Mhu92umJgY5efnuxzfv3+/UlNTFRUVJZvNptzc3B6vl52dLZvNpoyMjL60DwAAYCSvg93WrVuVkZGhpUuXqrKyUomJiUpJSVFtba3H+pqaGk2bNk2JiYmqrKzUkiVLtHDhQhUVFTlrTpw4oZiYGK1YsUJhYWE9zv+3v/1NBQUFGjdunLetAwAAGM3rYLdy5UrNmTNHc+fOVWxsrHJzcxUREaG8vDyP9fn5+Ro1apRyc3MVGxuruXPnavbs2crJyXHWTJw4UU899ZTuuOMOBQQEnHbuY8eOacaMGVq3bp2GDBnibesAAABG8yrYdXR0qKKiQsnJyS7jycnJ2r17t8dzysrK3OqnTp2q8vJydXZ2etXsggULdPPNN+vGG2/sVX17e7va2tpcNgAAAFN5FexaWlrU3d2t0NBQl/HQ0FA1NTV5PKepqcljfVdXl1paWno995YtW7R3715lZ2f3+pzs7GwFBwc7t4iIiF6fCwAAcKHp08MTNpvNZd+yLLexM9V7Gj+duro63Xfffdq0aZPsdnuv+8zKylJra6tzq6ur6/W5AAAAFxpfb4qHDh0qHx8ft9W55uZmt1W5U8LCwjzW+/r6KiQkpFfzVlRUqLm5WQ6HwznW3d2tt99+W6tXr1Z7e7t8fHzczgsICOjxb/YAAABM4tWKnb+/vxwOh0pKSlzGS0pKNHnyZI/nJCQkuNXv2LFD8fHx8vPz69W8N9xwg/bt26eqqirnFh8frxkzZqiqqspjqAMAALjYeLViJ0mZmZmaOXOm4uPjlZCQoIKCAtXW1io9PV3Stx9/1tfXa+PGjZKk9PR0rV69WpmZmZo3b57KyspUWFiozZs3O6/Z0dGhAwcOOH+ur69XVVWVAgMDNXr0aA0ePFhxcXEufQwaNEghISFu4wAAABcrr4NdWlqajhw5ouXLl6uxsVFxcXEqLi5WZGSkJKmxsdHlnXbR0dEqLi7WokWLtGbNGoWHh2vVqlVKTU111jQ0NGj8+PHO/ZycHOXk5CgpKUk7d+78HrcHAABw8bBZp55kuAi0tbUpODhYra2tCgoK6u92ABjkw/pW3fL7d/TKvdcpbkRwf7cDwCDe5Be+KxYAAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADOHb3w0AwJnUtBzX8fau/m6jR4eaj7n85/lsUICvoocO6u82AJwDBDsA57WaluP6ac7O/m6j1zK2VvV3C73y1gPXE+4AAxHsAJzXTq3U5aZdq9HDAvu5m9P7prNbn3/5tUYOGSi7n09/t3Nah5qPKWNr1Xm/Agqgbwh2AC4Io4cFKm5EcH+30aP4qP7uAMDFjocnAAAADEGwAwAAMATBDgAAwBB9CnZr165VdHS07Ha7HA6Hdu3a1WN9aWmpHA6H7Ha7YmJilJ+f73J8//79Sk1NVVRUlGw2m3Jzc92ukZeXp3HjxikoKEhBQUFKSEjQa6+91pf2AQAAjOR1sNu6dasyMjK0dOlSVVZWKjExUSkpKaqtrfVYX1NTo2nTpikxMVGVlZVasmSJFi5cqKKiImfNiRMnFBMToxUrVigsLMzjdUaOHKkVK1aovLxc5eXl+tnPfqbbbrtN+/fv9/YWAAAAjOR1sFu5cqXmzJmjuXPnKjY2Vrm5uYqIiFBeXp7H+vz8fI0aNUq5ubmKjY3V3LlzNXv2bOXk5DhrJk6cqKeeekp33HGHAgICPF7n1ltv1bRp03TFFVfoiiuu0OOPP67AwEDt2bPH21sAAAAwklfBrqOjQxUVFUpOTnYZT05O1u7duz2eU1ZW5lY/depUlZeXq7Oz08t2v9Xd3a0tW7bo+PHjSkhI6NM1AAAATOPVe+xaWlrU3d2t0NBQl/HQ0FA1NTV5PKepqcljfVdXl1paWjR8+PBez79v3z4lJCTom2++UWBgoF588UWNGTPmtPXt7e1qb2937re1tfV6LgAAgAtNnx6esNlsLvuWZbmNnane0/iZXHnllaqqqtKePXv07//+75o1a5YOHDhw2vrs7GwFBwc7t4iICK/mAwAAuJB4FeyGDh0qHx8ft9W55uZmt1W5U8LCwjzW+/r6KiQkxKtm/f39NXr0aMXHxys7O1vXXHONnnnmmdPWZ2VlqbW11bnV1dV5NR8AAMCFxKtg5+/vL4fDoZKSEpfxkpISTZ482eM5CQkJbvU7duxQfHy8/Pz8vGzXlWVZLh+1fldAQIDz9SinNgAAAFN5/V2xmZmZmjlzpuLj45WQkKCCggLV1tYqPT1d0rerZPX19dq4caMkKT09XatXr1ZmZqbmzZunsrIyFRYWavPmzc5rdnR0OD9S7ejoUH19vaqqqhQYGKjRo0dLkpYsWaKUlBRFRETo6NGj2rJli3bu3KnXX3/9e/8SAAAATOB1sEtLS9ORI0e0fPlyNTY2Ki4uTsXFxYqMjJQkNTY2urzTLjo6WsXFxVq0aJHWrFmj8PBwrVq1Sqmpqc6ahoYGjR8/3rmfk5OjnJwcJSUlaefOnZKk//3f/9XMmTPV2Nio4OBgjRs3Tq+//rpuuummvt47AACAUbwOdpI0f/58zZ8/3+OxDRs2uI0lJSVp7969p71eVFSU84GK0yksLPSqRwAAgIsN3xULAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhuhTsFu7dq2io6Nlt9vlcDi0a9euHutLS0vlcDhkt9sVExOj/Px8l+P79+9XamqqoqKiZLPZlJub63aN7OxsTZw4UYMHD9awYcM0ffp0ffTRR31pHwAAwEheB7utW7cqIyNDS5cuVWVlpRITE5WSkqLa2lqP9TU1NZo2bZoSExNVWVmpJUuWaOHChSoqKnLWnDhxQjExMVqxYoXCwsI8Xqe0tFQLFizQnj17VFJSoq6uLiUnJ+v48ePe3gIAAICRfL09YeXKlZozZ47mzp0rScrNzdX27duVl5en7Oxst/r8/HyNGjXKuQoXGxur8vJy5eTkKDU1VZI0ceJETZw4UZK0ePFij/O+/vrrLvvr16/XsGHDVFFRoZ/85Cfe3gYAAIBxvFqx6+joUEVFhZKTk13Gk5OTtXv3bo/nlJWVudVPnTpV5eXl6uzs9LLd/9Pa2ipJuvTSS09b097erra2NpcNAADAVF4Fu5aWFnV3dys0NNRlPDQ0VE1NTR7PaWpq8ljf1dWllpYWL9v9lmVZyszM1HXXXae4uLjT1mVnZys4ONi5RURE9Gk+AACAC0GfHp6w2Wwu+5ZluY2dqd7TeG/95je/0QcffKDNmzf3WJeVlaXW1lbnVldX16f5AAAALgRe/Y3d0KFD5ePj47Y619zc7LYqd0pYWJjHel9fX4WEhHjZrnTvvfdq27ZtevvttzVy5MgeawMCAhQQEOD1HAAAABcir1bs/P395XA4VFJS4jJeUlKiyZMnezwnISHBrX7Hjh2Kj4+Xn59fr+e2LEu/+c1v9D//8z/6y1/+oujoaG9aBwAAMJ7XT8VmZmZq5syZio+PV0JCggoKClRbW6v09HRJ3378WV9fr40bN0qS0tPTtXr1amVmZmrevHkqKytTYWGhy8eoHR0dOnDggPPn+vp6VVVVKTAwUKNHj5YkLViwQC+88IJeeuklDR482LkKGBwcrIEDB36/3wIAAIABvA52aWlpOnLkiJYvX67GxkbFxcWpuLhYkZGRkqTGxkaXd9pFR0eruLhYixYt0po1axQeHq5Vq1Y5X3UiSQ0NDRo/frxzPycnRzk5OUpKStLOnTslSXl5eZKk66+/3qWf9evX65577vH2NgAAAIzjdbCTpPnz52v+/Pkej23YsMFtLCkpSXv37j3t9aKiopwPVJzOmY4DAABc7PiuWAAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAM4dvfDQBAT9q7v9EAe71q2j7SAHtgf7dzwatpO6YB9nq1d38jKbi/2wFwlhHsAJzXGo5/pkHRv9eSv/Z3J+YYFC01HL9WDoX2dysAzjKCHYDzWvigSB2vuVfPpF2ry4exYvd9HW4+pvu2Vin8p5H93QqAc4BgB+C8FuBj18lvRig66EqNCeGjw+/r5DetOvnN3xXgY+/vVgCcAzw8AQAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiiT8Fu7dq1io6Olt1ul8Ph0K5du3qsLy0tlcPhkN1uV0xMjPLz812O79+/X6mpqYqKipLNZlNubq7bNd5++23deuutCg8Pl81m05///Oe+tA4AAGAsr4Pd1q1blZGRoaVLl6qyslKJiYlKSUlRbW2tx/qamhpNmzZNiYmJqqys1JIlS7Rw4UIVFRU5a06cOKGYmBitWLFCYWFhHq9z/PhxXXPNNVq9erW3LQMAAFwUfL09YeXKlZozZ47mzp0rScrNzdX27duVl5en7Oxst/r8/HyNGjXKuQoXGxur8vJy5eTkKDU1VZI0ceJETZw4UZK0ePFij/OmpKQoJSXF23YBAAAuGl6t2HV0dKiiokLJycku48nJydq9e7fHc8rKytzqp06dqvLycnV2dnrZLgAAAE7HqxW7lpYWdXd3KzQ01GU8NDRUTU1NHs9pamryWN/V1aWWlhYNHz7cy5Z7r729Xe3t7c79tra2czYXAABAf+vTwxM2m81l37Ist7Ez1XsaP9uys7MVHBzs3CIiIs7pfAAAAP3Jq2A3dOhQ+fj4uK3ONTc3u63KnRIWFuax3tfXVyEhIV62652srCy1trY6t7q6unM6HwAAQH/yKtj5+/vL4XCopKTEZbykpESTJ0/2eE5CQoJb/Y4dOxQfHy8/Pz8v2/VOQECAgoKCXDYAAABTef1UbGZmpmbOnKn4+HglJCSooKBAtbW1Sk9Pl/TtKll9fb02btwoSUpPT9fq1auVmZmpefPmqaysTIWFhdq8ebPzmh0dHTpw4IDz5/r6elVVVSkwMFCjR4+WJB07dkyHDh1ynlNTU6OqqipdeumlGjVqVN9/AwAAAIbwOtilpaXpyJEjWr58uRobGxUXF6fi4mJFRkZKkhobG13eaRcdHa3i4mItWrRIa9asUXh4uFatWuV81YkkNTQ0aPz48c79nJwc5eTkKCkpSTt37pQklZeX66c//amzJjMzU5I0a9YsbdiwwdvbAAAAMI7XwU6S5s+fr/nz53s85ilkJSUlae/evae9XlRUlPOBitO5/vrrz1gDAABwMeO7YgEAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEMQ7AAAAAxBsAMAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADEGwAwAAMATBDgAAwBAEOwAAAEP49ncDANCTrzu7JUkf1rf2cyc9+6azW59/+bVGDhkou59Pf7dzWoeaj/V3CwDOIYIdgPPa4f8/iCz+n3393IlZBgXwf/+AifhfNoDzWvLVYZKky4cFauB5vhKWsbVKuWnXavSwwP5up0eDAnwVPXRQf7cB4Bwg2AE4r106yF93/L9R/d1Gr40eFqi4EcH93QaAixQPTwAAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCH6FOzWrl2r6Oho2e12ORwO7dq1q8f60tJSORwO2e12xcTEKD8/3+X4/v37lZqaqqioKNlsNuXm5p6VeQEAAC4mXge7rVu3KiMjQ0uXLlVlZaUSExOVkpKi2tpaj/U1NTWaNm2aEhMTVVlZqSVLlmjhwoUqKipy1pw4cUIxMTFasWKFwsLCzsq8AAAAFxubZVmWNydMmjRJEyZMUF5ennMsNjZW06dPV3Z2tlv9Qw89pG3btqm6uto5lp6ervfff19lZWVu9VFRUcrIyFBGRsb3mteTtrY2BQcHq7W1VUFBQb06BwB648P6Vt3y+3f0yr3XKW5EcH+3A8Ag3uQXr1bsOjo6VFFRoeTkZJfx5ORk7d692+M5ZWVlbvVTp05VeXm5Ojs7z9m8ktTe3q62tjaXDQAAwFReBbuWlhZ1d3crNDTUZTw0NFRNTU0ez2lqavJY39XVpZaWlnM2ryRlZ2crODjYuUVERPRqPgAAgAtRnx6esNlsLvuWZbmNnane0/jZnjcrK0utra3Ora6uzqv5AAAALiS+3hQPHTpUPj4+bqtkzc3Nbqtpp4SFhXms9/X1VUhIyDmbV5ICAgIUEBDQqzkAAAAudF6t2Pn7+8vhcKikpMRlvKSkRJMnT/Z4TkJCglv9jh07FB8fLz8/v3M2LwAAwMXGqxU7ScrMzNTMmTMVHx+vhIQEFRQUqLa2Vunp6ZK+/fizvr5eGzdulPTtE7CrV69WZmam5s2bp7KyMhUWFmrz5s3Oa3Z0dOjAgQPOn+vr61VVVaXAwECNHj26V/MCAABc7LwOdmlpaTpy5IiWL1+uxsZGxcXFqbi4WJGRkZKkxsZGl3fLRUdHq7i4WIsWLdKaNWsUHh6uVatWKTU11VnT0NCg8ePHO/dzcnKUk5OjpKQk7dy5s1fzAgAAXOy8fo/dhYz32AE4V3iPHYBz5Zy9xw4AAADnL4IdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGAIgh0AAIAhCHYAAACGINgBAAAYgmAHAABgCIIdAACAIXz7ctLatWv11FNPqbGxUVdffbVyc3OVmJh42vrS0lJlZmZq//79Cg8P14MPPqj09HSXmqKiIi1btkyHDx/W5Zdfrscff1y/+MUvnMePHj2qZcuW6cUXX1Rzc7PGjx+vZ555RhMnTuzLLQC4iH3d0a3Dfz92Vq95qPmYy3+eTZdfFqiB/j5n/boAzON1sNu6dasyMjK0du1aTZkyRf/1X/+llJQUHThwQKNGjXKrr6mp0bRp0zRv3jxt2rRJ7777rubPn6/LLrtMqampkqSysjKlpaXp0Ucf1S9+8Qu9+OKL+pd/+Re98847mjRpkiRp7ty5+vDDD/WHP/xB4eHh2rRpk2688UYdOHBAI0aM+J6/BgAXk8N/P6Zbfv/OObl2xtaqs37NV+69TnEjgs/6dQGYx2ZZluXNCZMmTdKECROUl5fnHIuNjdX06dOVnZ3tVv/QQw9p27Ztqq6udo6lp6fr/fffV1lZmSQpLS1NbW1teu2115w1P//5zzVkyBBt3rxZX3/9tQYPHqyXXnpJN998s7Pm2muv1S233KLHHnusV723tbUpODhYra2tCgoK8ua2ARjkXKzYfdPZrc+//FojhwyU3e/srq6xYgdc3LzJL16t2HV0dKiiokKLFy92GU9OTtbu3bs9nlNWVqbk5GSXsalTp6qwsFCdnZ3y8/NTWVmZFi1a5FaTm5srSerq6lJ3d7fsdrtLzcCBA/XOO+fm37oBmGugv885WQGLjzrrlwQAr3j18ERLS4u6u7sVGhrqMh4aGqqmpiaP5zQ1NXms7+rqUktLS481p645ePBgJSQk6NFHH1VDQ4O6u7u1adMmvffee2psbDxtv+3t7Wpra3PZAAAATNWnp2JtNpvLvmVZbmNnqv/u+Jmu+Yc//EGWZWnEiBEKCAjQqlWrdNddd8nH5/QfT2RnZys4ONi5RUREnPnmAAAALlBeBbuhQ4fKx8fHbXWuubnZbcXtlLCwMI/1vr6+CgkJ6bHmH695+eWXq7S0VMeOHVNdXZ3++te/qrOzU9HR0aftNysrS62trc6trq7Om9sFAAC4oHgV7Pz9/eVwOFRSUuIyXlJSosmTJ3s8JyEhwa1+x44dio+Pl5+fX481nq45aNAgDR8+XF9++aW2b9+u22677bT9BgQEKCgoyGUDAAAwldevO8nMzNTMmTMVHx+vhIQEFRQUqLa21vleuqysLNXX12vjxo2Svn0CdvXq1crMzNS8efNUVlamwsJCbd682XnN++67Tz/5yU/05JNP6rbbbtNLL72kN954w+XBiO3bt8uyLF155ZU6dOiQfvvb3+rKK6/Uv/7rv37f3wEAAIARvA52aWlpOnLkiJYvX67GxkbFxcWpuLhYkZGRkqTGxkbV1tY666Ojo1VcXKxFixZpzZo1Cg8P16pVq5zvsJOkyZMna8uWLfqP//gPLVu2TJdffrm2bt3qfIedJLW2tiorK0uff/65Lr30UqWmpurxxx93rvoBAABc7Lx+j92FjPfYAQCAC403+YXvigUAADAEwQ4AAMAQBDsAAABDEOwAAAAMQbADAAAwBMEOAADAEAQ7AAAAQxDsAAAADOH1N09cyE69i7mtra2fOwEAAOidU7mlN98pcVEFu6NHj0qSIiIi+rkTAAAA7xw9elTBwcE91lxUXyl28uRJNTQ0aPDgwbLZbP3dDgCDtLW1KSIiQnV1dXxlIYCzyrIsHT16VOHh4RowoOe/oruogh0AnCt8FzWA8wEPTwAAABiCYAcAAGAIgh0AnAUBAQF6+OGHFRAQ0N+tALiI8Td2AAAAhmDFDgAAwBAEOwAAAEMQ7AAAAAxBsAOAs2Dnzp2y2Wz66quv+rsVABcxHp4AgLOgo6NDX3zxhUJDQ8/7b7a5/vrrde211yo3N7e/WwFwlrFiBwC90NnZ2eNxf39/hYWF9WuoO1OPAMxHsAPwgzt58qSefPJJjR49WgEBARo1apQef/xxSdK+ffv0s5/9TAMHDlRISIj+7d/+TceOHXOee88992j69OnKycnR8OHDFRISogULFjhDTVZWln784x+7zTlu3Dg9/PDDzv3169crNjZWdrtdV111ldauXes89umnn8pms+lPf/qTrr/+etntdm3atEmfffaZbr31Vg0ZMkSDBg3S1VdfreLiYkmeP4otKirS1VdfrYCAAEVFRenpp5926SkqKkpPPPGEZs+ercGDB2vUqFEqKCjo1e/wdD0eOXJEd955p0aOHKlLLrlEY8eO1ebNm11+f6WlpXrmmWdks9lks9n06aefSpIOHDigadOmKTAwUKGhoZo5c6ZaWlp61Q+A84QFAD+wBx980BoyZIi1YcMG69ChQ9auXbusdevWWcePH7fCw8Otf/7nf7b27dtnvfnmm1Z0dLQ1a9Ys57mzZs2ygoKCrPT0dKu6utp6+eWXrUsuucQqKCiwLMuy9u3bZ0myDh065Dznww8/tCRZH330kWVZllVQUGANHz7cKioqsj755BOrqKjIuvTSS60NGzZYlmVZNTU1liQrKirKWVNfX2/dfPPN1k033WR98MEH1uHDh62XX37ZKi0ttSzLst566y1LkvXll19almVZ5eXl1oABA6zly5dbH330kbV+/Xpr4MCB1vr16519RUZGWpdeeqm1Zs0a6+DBg1Z2drY1YMAAq7q6+oy/w9P1+Pnnn1tPPfWUVVlZaR0+fNhatWqV5ePjY+3Zs8eyLMv66quvrISEBGvevHlWY2Oj1djYaHV1dVkNDQ3W0KFDraysLKu6utrau3evddNNN1k//elP+/zfM4AfHsEOwA+qra3NCggIsNatW+d2rKCgwBoyZIh17Ngx59irr75qDRgwwGpqarIs69tgFxkZaXV1dTlrbr/9distLc25P27cOGv58uXO/aysLGvixInO/YiICOuFF15wmfvRRx+1EhISLMv6v9CUm5vrUjN27FjrkUce8Xhf3w12d911l3XTTTe51Pz2t7+1xowZ49yPjIy07r77buf+yZMnrWHDhll5eXke5/hHp+vRk2nTpln333+/cz8pKcm67777XGqWLVtmJScnu4zV1dW5BGIA5z8+igXwg6qurlZ7e7tuuOEGj8euueYaDRo0yDk2ZcoUnTx5Uh999JFz7Oqrr5aPj49zf/jw4Wpubnbuz5gxQ3/84x8lSZZlafPmzZoxY4Yk6e9//7vq6uo0Z84cBQYGOrfHHntMhw8fduknPj7eZX/hwoV67LHHNGXKFD388MP64IMPerzPKVOmuIxNmTJFBw8eVHd3t3Ns3Lhxzp9tNpvCwsJc7uVMvttjd3e3Hn/8cY0bN04hISEKDAzUjh07VFtb2+N1Kioq9NZbb7n8Tq666ipJcvu9ADh/+fZ3AwAuLgMHDjztMcuyTvvwwT+O+/n5uR07efKkc/+uu+7S4sWLtXfvXn399deqq6vTHXfcIUnOunXr1mnSpEku1/nHsCjJJWBK0ty5czV16lS9+uqr2rFjh7Kzs/X000/r3nvv7dW9WB5eQnCmezmT7/b49NNP6z//8z+Vm5ursWPHatCgQcrIyFBHR0eP1zl58qRuvfVWPfnkk27Hhg8f3ut+APQvgh2AH9SPfvQjDRw4UG+++abmzp3rcmzMmDF6/vnndfz4cWdgeffddzVgwABdccUVvZ5j5MiR+slPfqI//vGP+vrrr3XjjTcqNDRUkhQaGqoRI0bok08+ca7ieSMiIkLp6elKT09XVlaW1q1b5zHYjRkzRu+8847L2O7du3XFFVe4BcizadeuXbrtttt09913S/o2sB08eFCxsbHOGn9/f5dVQ0maMGGCioqKFBUVJV9f/tEAXKj4KBbAD8put+uhhx7Sgw8+qI0bN+rw4cPas2ePCgsLNWPGDNntds2aNUsffvih3nrrLd17772aOXOmM5j11owZM7Rlyxb993//tzPknPLII48oOztbzzzzjD7++GPt27dP69ev18qVK3u8ZkZGhrZv366amhrt3btXf/nLX1wC0z+6//779eabb+rRRx/Vxx9/rOeff16rV6/WAw884NV9eGv06NEqKSnR7t27VV1drV//+tdqampyqYmKitJ7772nTz/9VC0tLTp58qQWLFigL774Qnfeeaf++te/6pNPPtGOHTs0e/ZstxAI4PxFsAPwg1u2bJnuv/9+/e53v1NsbKzS0tLU3NysSy65RNu3b9cXX3yhiRMn6pe//KVuuOEGrV692us5br/9dh05ckQnTpzQ9OnTXY7NnTtXzz77rDZs2KCxY8cqKSlJGzZsUHR0dI/X7O7u1oIFCxQbG6uf//znuvLKK11ek/KPJkyYoD/96U/asmWL4uLi9Lvf/U7Lly/XPffc4/W9eGPZsmWaMGGCpk6dquuvv15hYWFu9//AAw/Ix8dHY8aM0WWXXaba2lqFh4fr3XffVXd3t6ZOnaq4uDjdd999Cg4O1oAB/KMCuFDwzRMAAACG4F/DAAAADEGwA4Dz0BNPPOHy6pF/3FJSUvq7PQDnKT6KBYDz0BdffKEvvvjC47GBAwdqxIgRP3BHAC4EBDsAAABD8FEsAACAIQh2AAAAhiDYAQAAGIJgBwAAYAiCHQAAgCEIdgAAAIYg2AEAABiCYAcAAGCI/w+3sxHnq2WOhgAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "conversion_rate.plot.box(whis=1.5)\n",
    "plt.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Results"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The results shows that the conversion rate of the linear regression is the worst (probably due to few samples available), thus we opt for the median rate which gave the best results in the chosen metrics."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 249,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "train, test = util_data.split_train_test(prices_ext, test_size=0.2, rng=rng)\n",
    "X_train, y_train = train[curr.lower()].values.reshape(-1, 1), train[\"eur\"].values\n",
    "X_test, y_test = test[curr.lower()].values.reshape(-1, 1), test[\"eur\"].values\n",
    "\n",
    "rate_train = train[\"eur\"] / train[curr.lower()]\n",
    "\n",
    "model_median = util_models.FixedLinearRegressor(\n",
    "    conversion_rate.median(), fit_intercept=False\n",
    ")\n",
    "model_linear = LinearRegression(fit_intercept=False)\n",
    "model_rate = util_models.FixedLinearRegressor(0.01268, fit_intercept=False)\n",
    "\n",
    "models = [model_median, model_fit, model_rate]\n",
    "names = [\"median\", \"linear regressor\", \"online_rate\"]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 250,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "median coefficients: [0.01148827]\n",
      "median RMSE: 34.01\n",
      "median r2-score: 0.71\n",
      "median MAE: 28.51\n",
      "\n",
      "linear regressor coefficients: [0.01086622]\n",
      "linear regressor RMSE: 38.09\n",
      "linear regressor r2-score: 0.64\n",
      "linear regressor MAE: 31.54\n",
      "\n",
      "online_rate coefficients: [0.01268]\n",
      "online_rate RMSE: 35.03\n",
      "online_rate r2-score: 0.70\n",
      "online_rate MAE: 31.87\n",
      "\n"
     ]
    }
   ],
   "source": [
    "y_preds = util_models.compare_models(X_train, y_train, X_test, y_test, models, names)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 252,
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# plot the data\n",
    "util_plot.plot_compare_models(\n",
    "    X_test, y_test, y_preds, names, f\"{curr}/EUR conversion rates\"\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "JFpUuz1CtSeV",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Continuous features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The display size measure the diagonal of the screen in inches, while the battery capacity is measured in mAh."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 294,
   "metadata": {
    "id": "d1nwXo_PtSeV",
    "outputId": "70e2dbdb-181f-47f9-e910-2dd29c39a079",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>display_size</th>\n",
       "      <th>battery</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>Removable Li-Ion battery</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>3.5 inches, 34.9 cm</td>\n",
       "      <td>Removable Li-Ion 1500 mAh battery</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3.2 inches, 30.5 cm (~46.3% screen-to-body ratio)</td>\n",
       "      <td>Removable Li-Po 1230 mAh battery</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2.8 inches, 24.3 cm (~43.7% screen-to-body ratio)</td>\n",
       "      <td>Removable Li-Ion 1500 mAh battery</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>6.3 inches, 98.0 cm (~81.3% screen-to-body ratio)</td>\n",
       "      <td>Non-removable Li-Po 3700 mAh battery</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                        display_size  \\\n",
       "0                                                NaN   \n",
       "1                                3.5 inches, 34.9 cm   \n",
       "2  3.2 inches, 30.5 cm (~46.3% screen-to-body ratio)   \n",
       "3  2.8 inches, 24.3 cm (~43.7% screen-to-body ratio)   \n",
       "4  6.3 inches, 98.0 cm (~81.3% screen-to-body ratio)   \n",
       "\n",
       "                                battery  \n",
       "0              Removable Li-Ion battery  \n",
       "1     Removable Li-Ion 1500 mAh battery  \n",
       "2      Removable Li-Po 1230 mAh battery  \n",
       "3     Removable Li-Ion 1500 mAh battery  \n",
       "4  Non-removable Li-Po 3700 mAh battery  "
      ]
     },
     "execution_count": 294,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"display_size\", \"battery\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "sAN_JUtctSeW",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Binary features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "These are *display_type*, *memory_card_slot*, *sound_loudspeaker*, *sound_3.5mm_jack*, *comms_wlan*, *comms_nfc*, *battery_charging* and *body*. \n",
    "\n",
    "These features are currently strings, but a more interesting representation is a binary one, so we can extract more relevant information from them."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "- display_type: the display has the OLED technology\n",
    "- comms_wlan: the technology of the wlan chipset, i.e. 2.4GHz or 5Ghz\n",
    "- battery_charging: extract whether the device has support for fast-charging\n",
    "- body: extract whether the device is waterproof\n",
    "- memory_card_slot: the device contains a slot for a microSD\n",
    "- sound_loudspeaker: the speaker is stereo or mono\n",
    "- sound_3.5mm_jack: the device has a 3.5mm jack\n",
    "- comms_nfc: the device supports the nfc"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Display type"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "2jK_qn0StSeW",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The *display_type* feature sometimes contains the keyword OLED, which depicts different display device technology, such as P-OLED, AMOLED or Super Retina. Otherwise the feature corresponds to a more general and worse technology, such as LCD or TFT.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 295,
   "metadata": {
    "id": "ACd3mNz3tSeW",
    "outputId": "c3bb91e1-3d11-4ea7-cb06-9986a923066a",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>display_type</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>Monochrome graphic</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>TFT resistive touchscreen, 65K colors</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>TFT capacitive touchscreen, 256K colors</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>TFT capacitive touchscreen, 256K colors</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>P-OLED capacitive touchscreen, 16M colors</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                display_type\n",
       "0                         Monochrome graphic\n",
       "1      TFT resistive touchscreen, 65K colors\n",
       "2    TFT capacitive touchscreen, 256K colors\n",
       "3    TFT capacitive touchscreen, 256K colors\n",
       "4  P-OLED capacitive touchscreen, 16M colors"
      ]
     },
     "execution_count": 295,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"display_type\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Wireless connectivity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "-9CJHWjrtSeW",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The *comms_wlan* feature contains the different [IEEE 802.11 standard](https://en.wikipedia.org/wiki/IEEE_802.11) supported. To distinguish whether the device supports the 5GHz or only the 2.4GHz depends on the standard adopted: the former is supported if at least one of \"a\" or \"ac\" generations is present in the feature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 291,
   "metadata": {
    "id": "m9t-vTcJtSeW",
    "outputId": "1ae44aab-8cd9-4358-f5e7-e208303c12d1",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>comms_wlan</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>No</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Wi-Fi 802.11b/g</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>Wi-Fi 802.11 b/g</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Wi-Fi 802.11 b/g</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Wi-Fi 802.11 a/b/g/n/ac, dual-band, Wi-Fi Dire...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                          comms_wlan\n",
       "0                                                 No\n",
       "1                                    Wi-Fi 802.11b/g\n",
       "2                                   Wi-Fi 802.11 b/g\n",
       "3                                   Wi-Fi 802.11 b/g\n",
       "4  Wi-Fi 802.11 a/b/g/n/ac, dual-band, Wi-Fi Dire..."
      ]
     },
     "execution_count": 291,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"comms_wlan\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Wireless charging"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ok2hGrAwtSeW",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "When the *battery_charging* feature contains the keyword wireless the device supports any type of wireless charging (QI), otherwise only a wired charging is possible.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 296,
   "metadata": {
    "id": "3holxTE_tSeX",
    "outputId": "08469266-0cc2-433f-9a35-f97837b0b045",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>battery_charging</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Fast charging 18W  USB Power Delivery 2.0  QI ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                    battery_charging\n",
       "0                                                NaN\n",
       "1                                                NaN\n",
       "2                                                NaN\n",
       "3                                                NaN\n",
       "4  Fast charging 18W  USB Power Delivery 2.0  QI ..."
      ]
     },
     "execution_count": 296,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"battery_charging\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Water resistence"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "YrxEmsyetSeX",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The *body* feature might contain the keywords \"IP6x\", \"splash\" or \"water\" when the device is waterproof, a characteristic important for expensive devices.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 297,
   "metadata": {
    "id": "BMOabbIetSeX",
    "outputId": "aa8ae056-f983-4d9c-8e93-7b28c53a1897",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>body</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Trackball</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>IP68 dust/water resistant (up to 1.5m for 30 m...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                body\n",
       "0                                                NaN\n",
       "1                                                NaN\n",
       "2                                                NaN\n",
       "3                                          Trackball\n",
       "4  IP68 dust/water resistant (up to 1.5m for 30 m..."
      ]
     },
     "execution_count": 297,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"body\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Miscellaneous"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "yD7KBU7ktSeX",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The remaining binary features usually contains the word \"Yes\" to describe the presence of the feature or in some case they explictly mention the feature.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 298,
   "metadata": {
    "id": "zuGpw8jntSeX",
    "outputId": "58e7fe04-4f24-4e94-ad6e-9eefc91f8b2a",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>memory_card_slot</th>\n",
       "      <th>sound_loudspeaker</th>\n",
       "      <th>sound_3.5mm_jack</th>\n",
       "      <th>comms_nfc</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>microSDHC (dedicated slot)</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>microSDHC (dedicated slot)</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>microSDHC (dedicated slot)</td>\n",
       "      <td>Yes</td>\n",
       "      <td>Yes</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>No</td>\n",
       "      <td>Yes, with stereo speakers</td>\n",
       "      <td>No</td>\n",
       "      <td>Yes</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "             memory_card_slot          sound_loudspeaker sound_3.5mm_jack  \\\n",
       "0                          No                         No               No   \n",
       "1  microSDHC (dedicated slot)                        Yes              Yes   \n",
       "2  microSDHC (dedicated slot)                        Yes              Yes   \n",
       "3  microSDHC (dedicated slot)                        Yes              Yes   \n",
       "4                          No  Yes, with stereo speakers               No   \n",
       "\n",
       "  comms_nfc  \n",
       "0       NaN  \n",
       "1       NaN  \n",
       "2       NaN  \n",
       "3       NaN  \n",
       "4       Yes  "
      ]
     },
     "execution_count": 298,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"memory_card_slot\", \"sound_loudspeaker\", \"sound_3.5mm_jack\", \"comms_nfc\"]].head()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "W4YalrrPtSeY",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Multi column features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The *display_resolution* feature contains the resolution of the display in pixels, both for the width and the height. If at least one of the two values is smaller than 1000, then the value is dropped because it's too small to be a smartphone display (smartwatch or an old smartphone).\n",
    "\n",
    "For the *memory_internal*, it represents the internal memory of the device in GB, both for the RAM and the ROM. When the device has at least one capacity expressed in MB, then the value is dropped because it's too small to be relevant.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 300,
   "metadata": {
    "id": "fr_fJt4atSeY",
    "outputId": "8c08784d-0ed3-4c01-9893-0efb83865fc6",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>display_resolution</th>\n",
       "      <th>memory_internal</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>6 lines</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>480 x 800 pixels, 5:3 ratio (~267 ppi density)</td>\n",
       "      <td>4GB 512MB RAM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>320 x 480 pixels, 3:2 ratio (~180 ppi density)</td>\n",
       "      <td>256MB RAM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>240 x 320 pixels, 4:3 ratio (~143 ppi density)</td>\n",
       "      <td>256MB RAM, 512MB</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1440 x 3040 pixels, 19:9 ratio (~537 ppi density)</td>\n",
       "      <td>64GB 6GB RAM, 128GB 6GB RAM</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                  display_resolution  \\\n",
       "0                                            6 lines   \n",
       "1     480 x 800 pixels, 5:3 ratio (~267 ppi density)   \n",
       "2     320 x 480 pixels, 3:2 ratio (~180 ppi density)   \n",
       "3     240 x 320 pixels, 4:3 ratio (~143 ppi density)   \n",
       "4  1440 x 3040 pixels, 19:9 ratio (~537 ppi density)   \n",
       "\n",
       "               memory_internal  \n",
       "0                          NaN  \n",
       "1                4GB 512MB RAM  \n",
       "2                    256MB RAM  \n",
       "3             256MB RAM, 512MB  \n",
       "4  64GB 6GB RAM, 128GB 6GB RAM  "
      ]
     },
     "execution_count": 300,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"display_resolution\", \"memory_internal\"]].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "y75iJrpFtSeY",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Network technology"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "The *network_technology* feature contains the different network technologies supported by the device. The most important ones are:\n",
    "\n",
    "- 5G: the device supports the 5G network\n",
    "- 4G: the device supports the 4G network, usually described as LTE.\n",
    "- 3G: the device supports the 3G network, usually described as UMTS or HSPA.\n",
    "\n",
    "The other technologies are less important, such as GSM, CDMA, EVDO or even devices with no network support are dropped.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 301,
   "metadata": {
    "id": "BnAfR5SwtSeY",
    "outputId": "df42a13b-f905-429e-cab6-561408808bfc",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>network_technology</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>GSM</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>GSM / HSPA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>GSM / HSPA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>GSM / HSPA</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>GSM / CDMA / HSPA / EVDO / LTE</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               network_technology\n",
       "0                             GSM\n",
       "1                      GSM / HSPA\n",
       "2                      GSM / HSPA\n",
       "3                      GSM / HSPA\n",
       "4  GSM / CDMA / HSPA / EVDO / LTE"
      ]
     },
     "execution_count": 301,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_copy[[\"network_technology\"]].head()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "ifN2lALYtSeZ",
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "#### Camera features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Each feature is split into different columns, each one describing a different setup of the camera(s) in the device. The columns are named as *main_camera_single*, *main_camera_dual*, etc. The number of columns is variable, depending on the number of cameras present in the device.\n",
    "\n",
    "Only one column for both \"main\" and \"selfie\" camera is filled (e.g. *main_camera_single* and *selfie_camera_dual* are filled), while the other columns are filled with NaN values. In this way the resolution of only the principal camera is considered, while the other ones are ignored. In addition, the number of cameras is extracted both for the main and the selfie cameras."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Main camera"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 302,
   "metadata": {
    "id": "ozhwN3KbtSeZ",
    "outputId": "c5db9ab0-70f8-452e-c160-93d32f466d51",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>main_camera_single</th>\n",
       "      <th>main_camera_dual</th>\n",
       "      <th>main_camera_quad</th>\n",
       "      <th>main_camera_triple</th>\n",
       "      <th>main_camera_five</th>\n",
       "      <th>main_camera_dual_or_triple</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5 MP, AF</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>5 MP, AF</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2 MP, AF</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>NaN</td>\n",
       "      <td>12.2 MP, f/1.7, 27mm (wide), 1/2.55\", 1.4&lt;c2&gt;&lt;...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  main_camera_single                                   main_camera_dual  \\\n",
       "0                NaN                                                NaN   \n",
       "1           5 MP, AF                                                NaN   \n",
       "2           5 MP, AF                                                NaN   \n",
       "3           2 MP, AF                                                NaN   \n",
       "4                NaN  12.2 MP, f/1.7, 27mm (wide), 1/2.55\", 1.4<c2><...   \n",
       "\n",
       "  main_camera_quad main_camera_triple main_camera_five  \\\n",
       "0              NaN                NaN              NaN   \n",
       "1              NaN                NaN              NaN   \n",
       "2              NaN                NaN              NaN   \n",
       "3              NaN                NaN              NaN   \n",
       "4              NaN                NaN              NaN   \n",
       "\n",
       "  main_camera_dual_or_triple  \n",
       "0                        NaN  \n",
       "1                        NaN  \n",
       "2                        NaN  \n",
       "3                        NaN  \n",
       "4                        NaN  "
      ]
     },
     "execution_count": 302,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "main_cols = [\"main_camera_single\", \"main_camera_dual\", \"main_camera_quad\", \"main_camera_triple\", \"main_camera_five\", \"main_camera_dual_or_triple\"]\n",
    "df_copy[main_cols].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Selfie camera"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 303,
   "metadata": {
    "id": "e1bSwYjLtSeZ",
    "outputId": "5db4a088-0d74-489e-9e57-930b0e73f02e",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "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>selfie_camera_single</th>\n",
       "      <th>selfie_camera_dual</th>\n",
       "      <th>selfie_camera_triple</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>NaN</td>\n",
       "      <td>8 MP, f/2.0, 22mm (wide), 1.22&lt;c2&gt;&lt;b5&gt;m, no AF...</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  selfie_camera_single                                 selfie_camera_dual  \\\n",
       "0                  NaN                                                NaN   \n",
       "1                  NaN                                                NaN   \n",
       "2                  NaN                                                NaN   \n",
       "3                  NaN                                                NaN   \n",
       "4                  NaN  8 MP, f/2.0, 22mm (wide), 1.22<c2><b5>m, no AF...   \n",
       "\n",
       "  selfie_camera_triple  \n",
       "0                  NaN  \n",
       "1                  NaN  \n",
       "2                  NaN  \n",
       "3                  NaN  \n",
       "4                  NaN  "
      ]
     },
     "execution_count": 303,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "selfie_cols = [\"selfie_camera_single\", \"selfie_camera_dual\", \"selfie_camera_triple\"]\n",
    "df_copy[selfie_cols].head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1cc9Y23uHMbq",
    "slideshow": {
     "slide_type": "slide"
    }
   },
   "source": [
    "## Preprocessing\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "We preprocess the data in order to make it suitable for the task. The preprocessing steps are the following:\n",
    "\n",
    "- Drop the columns which are not relevant for the task.\n",
    "- Process each feature in order to make it suitable for the task, as described in the previous section.\n",
    "- For some features, such as binary features, the NaN values are replaced to depict the absence of the feature. The samples where the NaN values in the features cannot be estimated are dropped.\n",
    "\n",
    "Many samples are dropped, but mostly because they are old devices, which are not relevant for the task, for example if the display resolution is smaller than 1000 pixels or if the internal memory is smaller than 1GB.\n",
    "\n",
    "The final dataset contains 2159 samples and 22 features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "2hRKCy2UtSeZ",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "filter_condition = [\n",
    "    (\"launch_announced\", lambda x: x >= 2014),\n",
    "    (\"display_size\", lambda x: (x >= 4) & (x <= 7.5)),\n",
    "    (\"display_height\", lambda x: x >= 512),\n",
    "    (\"display_width\", lambda x: x >= 512),\n",
    "    (\"num_main_camera\", lambda x: x <= 4),\n",
    "    (\"memory_rom_gb\", lambda x: x <= 256),\n",
    "    (\"memory_ram_gb\", lambda x: x <= 32),\n",
    "    (\"battery\", lambda x: x>= 1000),\n",
    "    (\"misc_price\", lambda x: x <= 2000),\n",
    "]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "Filtering"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xTolW29DtSea",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "source": [
    "The dataset is filtered to keep only the samples which satisfy the following conditions:\n",
    "\n",
    "- *launch_announced* greater or equal than 2014;\n",
    "- *display_size* between 4 inches and 7.5 inches;\n",
    "- *display_height* and *display_width* greater or equal than 512 pixels;\n",
    "- *num_main_camera* lower or equal than 4;\n",
    "- *battery* greater or equal than 1000 mAh\n",
    "- *misc_price* lower than 2000€.\n",
    "\n",
    "With these conditions the dataset is reduced but we can be sure that the samples are smartphones and the most recent."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "i7bm9FBbtSea",
    "outputId": "cff6338e-3da6-458f-c586-f257ba6b243a",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The dataframe has 2159 rows and 22 columns.\n"
     ]
    }
   ],
   "source": [
    "preprocess = util_preprocess.GSMArenaPreprocess(\"gsmarena.json\")\n",
    "df = preprocess(df_raw, filter_condition, verbose=False)\n",
    "\n",
    "print(f\"The dataframe has {df.shape[0]} rows and {df.shape[1]} columns.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "slideshow": {
     "slide_type": "subslide"
    }
   },
   "source": [
    "Final dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "lbNHuSiAtSea",
    "slideshow": {
     "slide_type": "fragment"
    }
   },
   "source": [
    "Instead of simply removing duplicates, we keep only one duplicated sample by considering the median price. This is done because the price is the target of the task and we want to avoid that the model is biased towards the most expensive device.\n",
    "\n",
    "Other techniques for detecting duplicates are possible, such as clustering or PCA, but the median price is a simple and effective way to remove duplicates.\n",
    "\n",
    "We end up with 1911 samples which can be used for the classification task."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "EIvThxMnuHF2",
    "outputId": "d18bbc68-b562-4f6c-b0c2-079afa0289c7",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The dataframe has 2159 rows and 22 columns.\n",
      "Number of duplicated rows: 182\n",
      "The dataframe after removing duplicates has 1729 rows\n",
      "The dataframe after adding duplicates has 1911 rows\n"
     ]
    }
   ],
   "source": [
    "# Show duplicated rows\n",
    "def get_dup_rows(df: pd.DataFrame, drop_cols):\n",
    "    dup_cols = df.columns.drop(drop_cols).to_list()\n",
    "    dup_rows = df[df.duplicated(subset=dup_cols, keep=False)].copy()\n",
    "    return dup_rows, dup_rows.groupby(dup_cols)\n",
    "\n",
    "\n",
    "df_duplicates, df_groups = get_dup_rows(df, [\"misc_price\", \"launch_announced\", \"oem_model\"])\n",
    "\n",
    "print(f\"The dataframe has {df.shape[0]} rows and {df.shape[1]} columns.\")\n",
    "print(f\"Number of duplicated rows: {df_groups.ngroups}\")\n",
    "\n",
    "# remove dupliates rows which are then substitued\n",
    "df = df.drop(df_duplicates.index).reset_index(drop=True)\n",
    "print(f\"The dataframe after removing duplicates has {df.shape[0]} rows\")\n",
    "\n",
    "# add duplicated rows by joining the rows and replacing with the median price\n",
    "for key, item in df_groups:\n",
    "    el = item.iloc[0, :].copy()\n",
    "    el.misc_price = item.misc_price.median()\n",
    "    el.launch_announced = item.launch_announced.max()\n",
    "\n",
    "    # define the oem and model\n",
    "    oem_model = item.oem_model.unique()\n",
    "    if len(oem_model) > 1:\n",
    "        oem_model = \" - \".join(oem_model)\n",
    "    else:\n",
    "        oem = oem[0]\n",
    "\n",
    "    el.oem_model = oem_model\n",
    "    \n",
    "    df = pd.concat([df, el.to_frame().T], axis=0)\n",
    "\n",
    "df = df.reset_index(drop=True)\n",
    "print(f\"The dataframe after adding duplicates has {df.shape[0]} rows\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "yYY2J5nLtSeb",
    "slideshow": {
     "slide_type": "skip"
    }
   },
   "outputs": [],
   "source": [
    "# save the dataframe\n",
    "df.to_csv(\"data/processed/gsm.csv\", index=True)"
   ]
  }
 ],
 "metadata": {
  "celltoolbar": "Slideshow",
  "colab": {
   "collapsed_sections": [
    "tETAOV0DJV6d"
   ],
   "provenance": []
  },
  "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": "3.9.16"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}