Entregable_Parte_2.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Entregable Parte 2.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/DiploDatos/AnalisisYCuracion/blob/master/Entregable_Parte_2.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zO4bRoxr2Apy"
      },
      "source": [
        "**Diplomatura en Ciencia de Datos, Aprendizaje Automático y sus Aplicaciones**\n",
        "\n",
        "**Exploración y Curación de Datos**\n",
        "\n",
        "*Edición 2021*\n",
        "\n",
        "----\n",
        "\n",
        "# Trabajo práctico entregable - parte 2\n",
        "\n",
        "\n",
        "En el ejercicio 1 de la parte 1 del entregable seleccionaron las filas y columnas relevantes al problema de predicción de precios de una propiedad. Además de ello, tuvieron que reducir el número de valores posibles para las variables categóricas utilizando información de dominio.\n",
        "\n",
        "En el ejercicio 2 de la parte 1 del entregable imputaron los valores faltantes de las columnas `Suburb` y las columnas obtenidas a partir del conjunto de datos `airbnb`.\n",
        "\n",
        "En esta notebook, **se utilizará resultado de dichas operaciones.**\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4udjxjk1EtVU"
      },
      "source": [
        "import matplotlib.pyplot as plt\n",
        "import numpy\n",
        "import pandas\n",
        "\n",
        "import seaborn\n",
        "seaborn.set_context('talk')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 227
        },
        "id": "_qeFN3GnEvMk",
        "outputId": "02e6e659-1be0-414f-a6a0-8f8298bb84d7"
      },
      "source": [
        "# Acá deberían leer el conjunto de datos que ya tienen.\n",
        "melb_df = pandas.read_csv(\n",
        "    'https://cs.famaf.unc.edu.ar/~mteruel/datasets/diplodatos/melb_data.csv')\n",
        "melb_df[:3]"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "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>Suburb</th>\n",
              "      <th>Address</th>\n",
              "      <th>Rooms</th>\n",
              "      <th>Type</th>\n",
              "      <th>Price</th>\n",
              "      <th>Method</th>\n",
              "      <th>SellerG</th>\n",
              "      <th>Date</th>\n",
              "      <th>Distance</th>\n",
              "      <th>Postcode</th>\n",
              "      <th>Bedroom2</th>\n",
              "      <th>Bathroom</th>\n",
              "      <th>Car</th>\n",
              "      <th>Landsize</th>\n",
              "      <th>BuildingArea</th>\n",
              "      <th>YearBuilt</th>\n",
              "      <th>CouncilArea</th>\n",
              "      <th>Lattitude</th>\n",
              "      <th>Longtitude</th>\n",
              "      <th>Regionname</th>\n",
              "      <th>Propertycount</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>Abbotsford</td>\n",
              "      <td>85 Turner St</td>\n",
              "      <td>2</td>\n",
              "      <td>h</td>\n",
              "      <td>1480000.0</td>\n",
              "      <td>S</td>\n",
              "      <td>Biggin</td>\n",
              "      <td>3/12/2016</td>\n",
              "      <td>2.5</td>\n",
              "      <td>3067.0</td>\n",
              "      <td>2.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>202.0</td>\n",
              "      <td>NaN</td>\n",
              "      <td>NaN</td>\n",
              "      <td>Yarra</td>\n",
              "      <td>-37.7996</td>\n",
              "      <td>144.9984</td>\n",
              "      <td>Northern Metropolitan</td>\n",
              "      <td>4019.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>Abbotsford</td>\n",
              "      <td>25 Bloomburg St</td>\n",
              "      <td>2</td>\n",
              "      <td>h</td>\n",
              "      <td>1035000.0</td>\n",
              "      <td>S</td>\n",
              "      <td>Biggin</td>\n",
              "      <td>4/02/2016</td>\n",
              "      <td>2.5</td>\n",
              "      <td>3067.0</td>\n",
              "      <td>2.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>156.0</td>\n",
              "      <td>79.0</td>\n",
              "      <td>1900.0</td>\n",
              "      <td>Yarra</td>\n",
              "      <td>-37.8079</td>\n",
              "      <td>144.9934</td>\n",
              "      <td>Northern Metropolitan</td>\n",
              "      <td>4019.0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>Abbotsford</td>\n",
              "      <td>5 Charles St</td>\n",
              "      <td>3</td>\n",
              "      <td>h</td>\n",
              "      <td>1465000.0</td>\n",
              "      <td>SP</td>\n",
              "      <td>Biggin</td>\n",
              "      <td>4/03/2017</td>\n",
              "      <td>2.5</td>\n",
              "      <td>3067.0</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>134.0</td>\n",
              "      <td>150.0</td>\n",
              "      <td>1900.0</td>\n",
              "      <td>Yarra</td>\n",
              "      <td>-37.8093</td>\n",
              "      <td>144.9944</td>\n",
              "      <td>Northern Metropolitan</td>\n",
              "      <td>4019.0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "       Suburb          Address  ...             Regionname Propertycount\n",
              "0  Abbotsford     85 Turner St  ...  Northern Metropolitan        4019.0\n",
              "1  Abbotsford  25 Bloomburg St  ...  Northern Metropolitan        4019.0\n",
              "2  Abbotsford     5 Charles St  ...  Northern Metropolitan        4019.0\n",
              "\n",
              "[3 rows x 21 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "s-mixICN22kA"
      },
      "source": [
        "## Ejercicio 1: Encoding\n",
        "\n",
        "1. Seleccionar todas las filas y columnas del conjunto de datos obtenido en la parte 1 del entregable, **excepto** `BuildingArea` y `YearBuilt`, que volveremos a imputar más adelante.\n",
        "\n",
        "2. Aplicar una codificación One-hot encoding a cada fila, tanto para variables numéricas como categóricas. Si lo consideran necesario, pueden volver a reducir el número de categorías únicas.\n",
        "\n",
        "Algunas opciones:\n",
        "  1. Utilizar `OneHotEncoder` junto con el parámetro `categories` para las variables categóricas y luego usar `numpy.hstack` para concatenar el resultado con las variables numéricas. \n",
        "  2. `DictVectorizer` con algunos pasos de pre-proceso previo.\n",
        "\n",
        "Recordar también que el atributo `pandas.DataFrame.values` permite acceder a la matriz de numpy subyacente a un DataFrame.\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K_Qk43aLIl4d"
      },
      "source": [
        "categorical_cols = ['Type']\n",
        "numerical_cols = ['Rooms']"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "oSNfW1BBJnV0",
        "outputId": "10ac472d-62ef-4e9b-f66f-48af6797b3bc"
      },
      "source": [
        "melb_df[categorical_cols].nunique()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Type    3\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XPDDOVHNNywn",
        "outputId": "4c497064-cca1-4344-b5a1-e8521ecb4cbc"
      },
      "source": [
        "# Check for nulls\n",
        "melb_df[categorical_cols].isna().sum()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Type    0\n",
              "dtype: int64"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ismngxPcfoWb"
      },
      "source": [
        "## Ejercicio 2: Imputación por KNN\n",
        "\n",
        "En el teórico se presentó el método `IterativeImputer` para imputar valores faltantes en variables numéricas. Sin embargo, los ejemplos presentados sólo utilizaban algunas variables numéricas presentes en el conjunto de datos. En este ejercicio, utilizaremos la matriz de datos codificada para imputar datos faltantes de manera más precisa.\n",
        "\n",
        "1. Agregue a la matriz obtenida en el punto anterior las columnas `YearBuilt` y `BuildingArea`.\n",
        "2. Aplique una instancia de `IterativeImputer` con un estimador `KNeighborsRegressor` para imputar los valores de las variables. ¿Es necesario estandarizar o escalar los datos previamente?\n",
        "3. Realice un gráfico mostrando la distribución de cada variable antes de ser imputada, y con ambos métodos de imputación."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "G4ClSr_JapCw",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "9468199f-071a-4514-c62b-e161a6113f18"
      },
      "source": [
        "from sklearn.experimental import enable_iterative_imputer\n",
        "from sklearn.neighbors import KNeighborsRegressor\n",
        "from sklearn.impute import IterativeImputer\n",
        "\n",
        "melb_data_mice = melb_df.copy(deep=True)\n",
        "\n",
        "mice_imputer = IterativeImputer(random_state=0, estimator=KNeighborsRegressor())\n",
        "melb_data_mice[['YearBuilt','BuildingArea']] = mice_imputer.fit_transform(\n",
        "    melb_data_mice[['YearBuilt', 'BuildingArea']])"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/sklearn/impute/_iterative.py:638: ConvergenceWarning: [IterativeImputer] Early stopping criterion not reached.\n",
            "  \" reached.\", ConvergenceWarning)\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ImjXQZUbVoKH"
      },
      "source": [
        "Ejemplo de gráfico comparando las distribuciones de datos obtenidas con cada método de imputación."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 355
        },
        "id": "PMK1ktqYQTJK",
        "outputId": "b3ec0023-25cd-41ad-b069-cc59de7c2313"
      },
      "source": [
        "mice_year_built = melb_data_mice.YearBuilt.to_frame()\n",
        "mice_year_built['Imputation'] = 'KNN over YearBuilt and BuildingArea'\n",
        "melb_year_build = melb_df.YearBuilt.dropna().to_frame()\n",
        "melb_year_build['Imputation'] = 'Original'\n",
        "data = pandas.concat([mice_year_built, melb_year_build])\n",
        "fig = plt.figure(figsize=(8, 5))\n",
        "g = seaborn.kdeplot(data=data, x='YearBuilt', hue='Imputation')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 576x360 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NBN7-5OIxjJW"
      },
      "source": [
        "## Ejercicio 3: Reducción de dimensionalidad.\n",
        "\n",
        "Utilizando la matriz obtenida en el ejercicio anterior:\n",
        "1. Aplique `PCA` para obtener $n$ componentes principales de la matriz, donde `n = min(20, X.shape[0])`. ¿Es necesario estandarizar o escalar los datos?\n",
        "2. Grafique la varianza capturada por los primeros $n$ componentes principales, para cada $n$.\n",
        "3. En base al gráfico, seleccione las primeras $m$ columnas de la matriz transformada para agregar como nuevas características al conjunto de datos."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "WrZTYmG_ZyDy"
      },
      "source": [
        "## Ejercicio 4: Composición del resultado\n",
        "\n",
        "Transformar nuevamente el conjunto de datos procesado en un `pandas.DataFrame` y guardarlo en un archivo.\n",
        "\n",
        "Para eso, será necesario recordar el nombre original de cada columna de la matriz, en el orden correcto. Tener en cuenta:\n",
        "1. El método `OneHotEncoder.get_feature_names` o el atributo `OneHotEncoder.categories_` permiten obtener una lista con los valores de la categoría que le corresponde a cada índice de la matriz.\n",
        "2. Ninguno de los métodos aplicados intercambia de lugar las columnas o las filas de la matriz."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DfchYPgTxvQ4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 287
        },
        "outputId": "c0bd0b28-1cb8-419c-e799-631045735021"
      },
      "source": [
        "## Small example\n",
        "from sklearn.decomposition import PCA\n",
        "from sklearn.preprocessing import OneHotEncoder\n",
        "\n",
        "## If we process our data with the following steps:\n",
        "categorical_cols = ['Type', 'Regionname']\n",
        "numerical_cols = ['Rooms', 'Distance']\n",
        "new_columns = []\n",
        "\n",
        "# Step 1: encode categorical columns\n",
        "encoder = OneHotEncoder(handle_unknown='ignore', sparse=False)\n",
        "X_cat = encoder.fit_transform(melb_df[categorical_cols])\n",
        "for col, col_values in zip(categorical_cols, encoder.categories_):\n",
        "  for col_value in col_values:\n",
        "    new_columns.append('{}={}'.format(col, col_value))\n",
        "print(\"Matrix has shape {}, with columns: {}\".format(X_cat.shape, new_columns))\n",
        "\n",
        "# Step 2: Append the numerical columns\n",
        "X = numpy.hstack([X_cat, melb_df[numerical_cols].values])\n",
        "new_columns.extend(numerical_cols)\n",
        "print(\"Matrix has shape {}, with columns: {}\".format(X_cat.shape, new_columns))\n",
        "\n",
        "# Step 3: Append some new features, like PCA\n",
        "pca = PCA(n_components=2)\n",
        "pca_dummy_features = pca.fit_transform(X)\n",
        "X_pca = numpy.hstack([X, pca_dummy_features])\n",
        "new_columns.extend(['pca1', 'pca2'])\n",
        "\n",
        "## Re-build dataframe\n",
        "processed_melb_df = pandas.DataFrame(data=X_pca, columns=new_columns)\n",
        "processed_melb_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Matrix has shape (13580, 11), with columns: ['Type=h', 'Type=t', 'Type=u', 'Regionname=Eastern Metropolitan', 'Regionname=Eastern Victoria', 'Regionname=Northern Metropolitan', 'Regionname=Northern Victoria', 'Regionname=South-Eastern Metropolitan', 'Regionname=Southern Metropolitan', 'Regionname=Western Metropolitan', 'Regionname=Western Victoria']\n",
            "Matrix has shape (13580, 11), with columns: ['Type=h', 'Type=t', 'Type=u', 'Regionname=Eastern Metropolitan', 'Regionname=Eastern Victoria', 'Regionname=Northern Metropolitan', 'Regionname=Northern Victoria', 'Regionname=South-Eastern Metropolitan', 'Regionname=Southern Metropolitan', 'Regionname=Western Metropolitan', 'Regionname=Western Victoria', 'Rooms', 'Distance']\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "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>Type=h</th>\n",
              "      <th>Type=t</th>\n",
              "      <th>Type=u</th>\n",
              "      <th>Regionname=Eastern Metropolitan</th>\n",
              "      <th>Regionname=Eastern Victoria</th>\n",
              "      <th>Regionname=Northern Metropolitan</th>\n",
              "      <th>Regionname=Northern Victoria</th>\n",
              "      <th>Regionname=South-Eastern Metropolitan</th>\n",
              "      <th>Regionname=Southern Metropolitan</th>\n",
              "      <th>Regionname=Western Metropolitan</th>\n",
              "      <th>Regionname=Western Victoria</th>\n",
              "      <th>Rooms</th>\n",
              "      <th>Distance</th>\n",
              "      <th>pca1</th>\n",
              "      <th>pca2</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>2.0</td>\n",
              "      <td>2.5</td>\n",
              "      <td>-7.669418</td>\n",
              "      <td>-0.292703</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>2.0</td>\n",
              "      <td>2.5</td>\n",
              "      <td>-7.669418</td>\n",
              "      <td>-0.292703</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.5</td>\n",
              "      <td>-7.620201</td>\n",
              "      <td>0.619633</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>3.0</td>\n",
              "      <td>2.5</td>\n",
              "      <td>-7.620201</td>\n",
              "      <td>0.619633</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>1.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.0</td>\n",
              "      <td>4.0</td>\n",
              "      <td>2.5</td>\n",
              "      <td>-7.570984</td>\n",
              "      <td>1.531969</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   Type=h  Type=t  Type=u  ...  Distance      pca1      pca2\n",
              "0     1.0     0.0     0.0  ...       2.5 -7.669418 -0.292703\n",
              "1     1.0     0.0     0.0  ...       2.5 -7.669418 -0.292703\n",
              "2     1.0     0.0     0.0  ...       2.5 -7.620201  0.619633\n",
              "3     1.0     0.0     0.0  ...       2.5 -7.620201  0.619633\n",
              "4     1.0     0.0     0.0  ...       2.5 -7.570984  1.531969\n",
              "\n",
              "[5 rows x 15 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 147
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "mVBLFc8PhRtW"
      },
      "source": [
        "## Ejercicio 5: Documentación\n",
        "\n",
        "En un documento `.pdf` o `.md` realizar un reporte de las operaciones que realizaron para obtener el conjunto de datos final. Se debe incluir:\n",
        "  1. Criterios de exclusión (o inclusión) de filas\n",
        "  2. Interpretación de las columnas presentes\n",
        "  2. Todas las transofrmaciones realizadas\n",
        "\n",
        "Este documento es de uso técnico exclusivamente, y su objetivo es permitir que otres desarrolladores puedan reproducir los mismos pasos y obtener el mismo resultado. Debe ser detallado pero consiso. Por ejemplo:\n",
        "\n",
        "```\n",
        "  ## Criterios de exclusión de ejemplos\n",
        "  1. Se eliminan ejemplos donde el año de construcción es previo a 1900\n",
        "\n",
        "  ## Características seleccionadas\n",
        "  ### Características categóricas\n",
        "  1. Type: tipo de propiedad. 3 valores posibles\n",
        "  2. ...\n",
        "  Todas las características categóricas fueron codificadas con un\n",
        "  método OneHotEncoding utilizando como máximo sus 30 valores más \n",
        "  frecuentes.\n",
        "  \n",
        "  ### Características numéricas\n",
        "  1. Rooms: Cantidad de habitaciones\n",
        "  2. Distance: Distancia al centro de la ciudad.\n",
        "  3. airbnb_mean_price: Se agrega el precio promedio diario de \n",
        "     publicaciones de la plataforma AirBnB en el mismo código \n",
        "     postal. [Link al repositorio con datos externos].\n",
        "\n",
        "  ### Transformaciones:\n",
        "  1. Todas las características numéricas fueron estandarizadas.\n",
        "  2. La columna `Suburb` fue imputada utilizando el método ...\n",
        "  3. Las columnas `YearBuilt` y ... fueron imputadas utilizando el \n",
        "     algoritmo ...\n",
        "  4. ...\n",
        "\n",
        "  ### Datos aumentados\n",
        "  1. Se agregan las 5 primeras columnas obtenidas a través del\n",
        "     método de PCA, aplicado sobre el conjunto de datos\n",
        "     totalmente procesado.\n",
        "```\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Bq7nuPg2HIYx"
      },
      "source": [
        ""
      ]
    }
  ]
}