deeplearning_timeseries_analysis_template.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "deeplearning-timeseries-analysis-template.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "hjq-ad6NgSsU"
      ]
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "source": [
        "Ref: https://github.com/PacktPublishing/Python-for-Finance-Cookbook/tree/master/Chapter%2010"
      ],
      "metadata": {
        "id": "Qp13NUhL7Mlg"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Tabular Classification"
      ],
      "metadata": {
        "id": "6H1UAoPq-5VE"
      }
    },
    {
      "cell_type": "code",
      "execution_count": 1,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 592
        },
        "id": "5annsV3D7HRw",
        "outputId": "6efb9ad7-2dee-4d16-a6e9-261fd9f08e58"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-d45c05ee-797d-463c-bbf7-5b1331f5769b\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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>limit_bal</th>\n",
              "      <th>sex</th>\n",
              "      <th>education</th>\n",
              "      <th>marriage</th>\n",
              "      <th>age</th>\n",
              "      <th>payment_status_sep</th>\n",
              "      <th>payment_status_aug</th>\n",
              "      <th>payment_status_jul</th>\n",
              "      <th>payment_status_jun</th>\n",
              "      <th>payment_status_may</th>\n",
              "      <th>payment_status_apr</th>\n",
              "      <th>bill_statement_sep</th>\n",
              "      <th>bill_statement_aug</th>\n",
              "      <th>bill_statement_jul</th>\n",
              "      <th>bill_statement_jun</th>\n",
              "      <th>bill_statement_may</th>\n",
              "      <th>bill_statement_apr</th>\n",
              "      <th>previous_payment_sep</th>\n",
              "      <th>previous_payment_aug</th>\n",
              "      <th>previous_payment_jul</th>\n",
              "      <th>previous_payment_jun</th>\n",
              "      <th>previous_payment_may</th>\n",
              "      <th>previous_payment_apr</th>\n",
              "      <th>default_payment_next_month</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>20000</td>\n",
              "      <td>Female</td>\n",
              "      <td>University</td>\n",
              "      <td>Married</td>\n",
              "      <td>24.0</td>\n",
              "      <td>Payment delayed 2 months</td>\n",
              "      <td>Payment delayed 2 months</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>3913</td>\n",
              "      <td>3102</td>\n",
              "      <td>689</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>689</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>120000</td>\n",
              "      <td>Female</td>\n",
              "      <td>University</td>\n",
              "      <td>Single</td>\n",
              "      <td>26.0</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Payment delayed 2 months</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Payment delayed 2 months</td>\n",
              "      <td>2682</td>\n",
              "      <td>1725</td>\n",
              "      <td>2682</td>\n",
              "      <td>3272</td>\n",
              "      <td>3455</td>\n",
              "      <td>3261</td>\n",
              "      <td>0</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>0</td>\n",
              "      <td>2000</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>90000</td>\n",
              "      <td>Female</td>\n",
              "      <td>University</td>\n",
              "      <td>Single</td>\n",
              "      <td>34.0</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>29239</td>\n",
              "      <td>14027</td>\n",
              "      <td>13559</td>\n",
              "      <td>14331</td>\n",
              "      <td>14948</td>\n",
              "      <td>15549</td>\n",
              "      <td>1518</td>\n",
              "      <td>1500</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>5000</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>50000</td>\n",
              "      <td>Female</td>\n",
              "      <td>University</td>\n",
              "      <td>Married</td>\n",
              "      <td>37.0</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>46990</td>\n",
              "      <td>48233</td>\n",
              "      <td>49291</td>\n",
              "      <td>28314</td>\n",
              "      <td>28959</td>\n",
              "      <td>29547</td>\n",
              "      <td>2000</td>\n",
              "      <td>2019</td>\n",
              "      <td>1200</td>\n",
              "      <td>1100</td>\n",
              "      <td>1069</td>\n",
              "      <td>1000</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>50000</td>\n",
              "      <td>Male</td>\n",
              "      <td>University</td>\n",
              "      <td>Married</td>\n",
              "      <td>57.0</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>8617</td>\n",
              "      <td>5670</td>\n",
              "      <td>35835</td>\n",
              "      <td>20940</td>\n",
              "      <td>19146</td>\n",
              "      <td>19131</td>\n",
              "      <td>2000</td>\n",
              "      <td>36681</td>\n",
              "      <td>10000</td>\n",
              "      <td>9000</td>\n",
              "      <td>689</td>\n",
              "      <td>679</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>...</th>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "      <td>...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>29995</th>\n",
              "      <td>220000</td>\n",
              "      <td>NaN</td>\n",
              "      <td>High school</td>\n",
              "      <td>Married</td>\n",
              "      <td>39.0</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>188948</td>\n",
              "      <td>192815</td>\n",
              "      <td>208365</td>\n",
              "      <td>88004</td>\n",
              "      <td>31237</td>\n",
              "      <td>15980</td>\n",
              "      <td>8500</td>\n",
              "      <td>20000</td>\n",
              "      <td>5003</td>\n",
              "      <td>3047</td>\n",
              "      <td>5000</td>\n",
              "      <td>1000</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>29996</th>\n",
              "      <td>150000</td>\n",
              "      <td>Male</td>\n",
              "      <td>High school</td>\n",
              "      <td>Single</td>\n",
              "      <td>43.0</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>1683</td>\n",
              "      <td>1828</td>\n",
              "      <td>3502</td>\n",
              "      <td>8979</td>\n",
              "      <td>5190</td>\n",
              "      <td>0</td>\n",
              "      <td>1837</td>\n",
              "      <td>3526</td>\n",
              "      <td>8998</td>\n",
              "      <td>129</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>29997</th>\n",
              "      <td>30000</td>\n",
              "      <td>Male</td>\n",
              "      <td>University</td>\n",
              "      <td>Single</td>\n",
              "      <td>37.0</td>\n",
              "      <td>Payment delayed 4 months</td>\n",
              "      <td>Payment delayed 3 months</td>\n",
              "      <td>Payment delayed 2 months</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>3565</td>\n",
              "      <td>3356</td>\n",
              "      <td>2758</td>\n",
              "      <td>20878</td>\n",
              "      <td>20582</td>\n",
              "      <td>19357</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "      <td>22000</td>\n",
              "      <td>4200</td>\n",
              "      <td>2000</td>\n",
              "      <td>3100</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>29998</th>\n",
              "      <td>80000</td>\n",
              "      <td>Male</td>\n",
              "      <td>High school</td>\n",
              "      <td>Married</td>\n",
              "      <td>41.0</td>\n",
              "      <td>Payment delayed 1 month</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Payed duly</td>\n",
              "      <td>-1645</td>\n",
              "      <td>78379</td>\n",
              "      <td>76304</td>\n",
              "      <td>52774</td>\n",
              "      <td>11855</td>\n",
              "      <td>48944</td>\n",
              "      <td>85900</td>\n",
              "      <td>3409</td>\n",
              "      <td>1178</td>\n",
              "      <td>1926</td>\n",
              "      <td>52964</td>\n",
              "      <td>1804</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>29999</th>\n",
              "      <td>50000</td>\n",
              "      <td>Male</td>\n",
              "      <td>University</td>\n",
              "      <td>Married</td>\n",
              "      <td>46.0</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>47929</td>\n",
              "      <td>48905</td>\n",
              "      <td>49764</td>\n",
              "      <td>36535</td>\n",
              "      <td>32428</td>\n",
              "      <td>15313</td>\n",
              "      <td>2078</td>\n",
              "      <td>1800</td>\n",
              "      <td>1430</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>1000</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>30000 rows × 24 columns</p>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-d45c05ee-797d-463c-bbf7-5b1331f5769b')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-d45c05ee-797d-463c-bbf7-5b1331f5769b button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-d45c05ee-797d-463c-bbf7-5b1331f5769b');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ],
            "text/plain": [
              "       limit_bal     sex  ... previous_payment_apr default_payment_next_month\n",
              "0          20000  Female  ...                    0                          1\n",
              "1         120000  Female  ...                 2000                          1\n",
              "2          90000  Female  ...                 5000                          0\n",
              "3          50000  Female  ...                 1000                          0\n",
              "4          50000    Male  ...                  679                          0\n",
              "...          ...     ...  ...                  ...                        ...\n",
              "29995     220000     NaN  ...                 1000                          0\n",
              "29996     150000    Male  ...                    0                          0\n",
              "29997      30000    Male  ...                 3100                          1\n",
              "29998      80000    Male  ...                 1804                          1\n",
              "29999      50000    Male  ...                 1000                          1\n",
              "\n",
              "[30000 rows x 24 columns]"
            ]
          },
          "metadata": {},
          "execution_count": 1
        }
      ],
      "source": [
        "# Load Data\n",
        "from fastai import *\n",
        "from fastai.tabular import *\n",
        "import pandas as pd\n",
        "import numpy as np\n",
        "\n",
        "df = pd.read_csv(\"https://raw.githubusercontent.com/PacktPublishing/Python-for-Finance-Cookbook/master/Datasets/credit_card_default.csv\", index_col=0, na_values='')\n",
        "df"
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Identify target and features\n",
        "DEP_VAR = 'default_payment_next_month'\n",
        "\n",
        "num_features = list(df.select_dtypes('number').columns)\n",
        "num_features.remove(DEP_VAR)\n",
        "cat_features = list(df.select_dtypes('object').columns)\n",
        "\n",
        "# list of preprocessing\n",
        "preprocessing = [FillMissing, Categorify, Normalize]"
      ],
      "metadata": {
        "id": "5gMtI16a_Bzq"
      },
      "execution_count": 2,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# Create dataframe\n",
        "data = (TabularList.from_df(df, \n",
        "                            cat_names=cat_features,\n",
        "                            cont_names=num_features, \n",
        "                            procs=preprocessing)\n",
        "                   .split_by_rand_pct(valid_pct=0.2, seed=42)\n",
        "                   .label_from_df(cols=DEP_VAR)\n",
        "                   .databunch())\n",
        "data"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "hURiIhej_I84",
        "outputId": "cd6b4e50-cafb-432e-9584-6ac08317f209"
      },
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "TabularDataBunch;\n",
              "\n",
              "Train: LabelList (24000 items)\n",
              "x: TabularList\n",
              "sex Female; education University; marriage Married; payment_status_sep Payment delayed 2 months; payment_status_aug Payment delayed 2 months; payment_status_jul Payed duly; payment_status_jun Payed duly; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -1.1348; age -1.2474; bill_statement_sep -0.6405; bill_statement_aug -0.6459; bill_statement_jul -0.6651; bill_statement_jun -0.6714; bill_statement_may -0.6622; bill_statement_apr -0.6539; previous_payment_sep -0.3319; previous_payment_aug -0.2171; previous_payment_jul -0.2882; previous_payment_jun -0.2994; previous_payment_may -0.3085; previous_payment_apr -0.2955; ,sex Female; education University; marriage Single; payment_status_sep Payed duly; payment_status_aug Payment delayed 2 months; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Payment delayed 2 months; age_na False; limit_bal -0.3640; age -1.0303; bill_statement_sep -0.6573; bill_statement_aug -0.6654; bill_statement_jul -0.6363; bill_statement_jun -0.6203; bill_statement_may -0.6052; bill_statement_apr -0.5989; previous_payment_sep -0.3319; previous_payment_aug -0.2043; previous_payment_jul -0.2334; previous_payment_jun -0.2380; previous_payment_may -0.3085; previous_payment_apr -0.1810; ,sex Female; education University; marriage Single; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -0.5953; age -0.1618; bill_statement_sep -0.2955; bill_statement_aug -0.4919; bill_statement_jul -0.4793; bill_statement_jun -0.4477; bill_statement_may -0.4156; bill_statement_apr -0.3916; previous_payment_sep -0.2431; previous_payment_aug -0.1837; previous_payment_jul -0.2334; previous_payment_jun -0.2380; previous_payment_may -0.2445; previous_payment_apr -0.0091; ,sex Female; education University; marriage Married; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -0.9036; age 0.1638; bill_statement_sep -0.0536; bill_statement_aug -0.0096; bill_statement_jul 0.0366; bill_statement_jun -0.2294; bill_statement_may -0.1846; bill_statement_apr -0.1555; previous_payment_sep -0.2149; previous_payment_aug -0.1623; previous_payment_jul -0.2225; previous_payment_jun -0.2319; previous_payment_may -0.2401; previous_payment_apr -0.2382; ,sex Male; education University; marriage Married; payment_status_sep Payed duly; payment_status_aug Unknown; payment_status_jul Payed duly; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -0.9036; age 2.3349; bill_statement_sep -0.5764; bill_statement_aug -0.6097; bill_statement_jul -0.1577; bill_statement_jun -0.3445; bill_statement_may -0.3464; bill_statement_apr -0.3312; previous_payment_sep -0.2149; previous_payment_aug 1.2650; previous_payment_jul 0.2599; previous_payment_jun 0.2526; previous_payment_may -0.2644; previous_payment_apr -0.2566; \n",
              "y: CategoryList\n",
              "1,1,0,0,0\n",
              "Path: .;\n",
              "\n",
              "Valid: LabelList (6000 items)\n",
              "x: TabularList\n",
              "sex #na#; education University; marriage Single; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -1.0577; age -1.1388; bill_statement_sep -0.5731; bill_statement_aug -0.5478; bill_statement_jul -0.5078; bill_statement_jun -0.4750; bill_statement_may -0.4360; bill_statement_apr -0.4038; previous_payment_sep -0.2441; previous_payment_aug -0.1631; previous_payment_jul -0.2060; previous_payment_jun -0.2074; previous_payment_may -0.2125; previous_payment_apr -0.1810; ,sex #na#; education Graduate school; marriage Single; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -0.1328; age -1.0303; bill_statement_sep 1.1690; bill_statement_aug 1.0820; bill_statement_jul 1.0096; bill_statement_jun 0.9141; bill_statement_may 0.6199; bill_statement_apr 0.6494; previous_payment_sep -0.0694; previous_payment_aug -0.0711; previous_payment_jul -0.1150; previous_payment_jun -0.1370; previous_payment_may -0.1376; previous_payment_apr -0.1426; ,sex #na#; education High school; marriage Married; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Unknown; age_na False; limit_bal -0.7494; age -0.3789; bill_statement_sep 0.2615; bill_statement_aug 0.2843; bill_statement_jul 0.3143; bill_statement_jun 0.4174; bill_statement_may 0.4941; bill_statement_apr 0.5305; previous_payment_sep -0.1896; previous_payment_aug -0.1173; previous_payment_jul -0.1238; previous_payment_jun -0.1498; previous_payment_may -0.1485; previous_payment_apr -0.1492; ,sex #na#; education High school; marriage Single; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Payed duly; age_na False; limit_bal -0.2869; age 1.4665; bill_statement_sep -0.4121; bill_statement_aug -0.4224; bill_statement_jul -0.4416; bill_statement_jun -0.4076; bill_statement_may -0.4769; bill_statement_apr -0.5367; previous_payment_sep -0.2377; previous_payment_aug -0.1710; previous_payment_jul 0.0962; previous_payment_jun -0.2977; previous_payment_may 0.1403; previous_payment_apr -0.0430; ,sex #na#; education University; marriage Single; payment_status_sep Unknown; payment_status_aug Unknown; payment_status_jul Unknown; payment_status_jun Unknown; payment_status_may Unknown; payment_status_apr Payment delayed 2 months; age_na False; limit_bal -0.9036; age 0.0553; bill_statement_sep 0.5899; bill_statement_aug -0.0180; bill_statement_jul -0.0635; bill_statement_jun -0.3658; bill_statement_may -0.3275; bill_statement_apr -0.3260; previous_payment_sep -0.2149; previous_payment_aug -0.1837; previous_payment_jul -0.2334; previous_payment_jun -0.1890; previous_payment_may -0.3085; previous_payment_apr -0.2382; \n",
              "y: CategoryList\n",
              "0,0,0,0,1\n",
              "Path: .;\n",
              "\n",
              "Test: None"
            ]
          },
          "metadata": {},
          "execution_count": 3
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Define Neural Net Learner\n",
        "learn = tabular_learner(data, layers=[1000,500], \n",
        "                        ps=[0.001,0.01], \n",
        "                        emb_drop=0.04, \n",
        "                        metrics=[Recall(), \n",
        "                                 FBeta(beta=1),\n",
        "                                 FBeta(beta=5)])\n",
        "\n",
        "# Find Learning Rate\n",
        "learn.lr_find()\n",
        "learn.recorder.plot(suggestion=True)\n",
        "plt.show()\n",
        "\n",
        "# Train Model\n",
        "learn.fit(epochs=25, lr=1e-6, wd=0.2)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "id": "EknXboZn_M0S",
        "outputId": "2f22e1b9-d8d8-4242-daf2-e191120d4338"
      },
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='0' class='' max='1' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      0.00% [0/1 00:00<00:00]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>recall</th>\n",
              "      <th>f_beta</th>\n",
              "      <th>f_beta</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='85' class='' max='375' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      22.67% [85/375 00:04<00:14 1.0838]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
            "Min numerical gradient: 7.59E-03\n",
            "Min loss divided by 10: 3.31E-05\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>recall</th>\n",
              "      <th>f_beta</th>\n",
              "      <th>f_beta</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>0.684715</td>\n",
              "      <td>0.656842</td>\n",
              "      <td>0.583397</td>\n",
              "      <td>0.385797</td>\n",
              "      <td>0.561283</td>\n",
              "      <td>00:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.678590</td>\n",
              "      <td>0.652812</td>\n",
              "      <td>0.583397</td>\n",
              "      <td>0.392619</td>\n",
              "      <td>0.562376</td>\n",
              "      <td>00:09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.671350</td>\n",
              "      <td>0.646393</td>\n",
              "      <td>0.600914</td>\n",
              "      <td>0.404304</td>\n",
              "      <td>0.579246</td>\n",
              "      <td>00:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.660906</td>\n",
              "      <td>0.639132</td>\n",
              "      <td>0.536938</td>\n",
              "      <td>0.395290</td>\n",
              "      <td>0.522535</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.662786</td>\n",
              "      <td>0.625267</td>\n",
              "      <td>0.448591</td>\n",
              "      <td>0.380123</td>\n",
              "      <td>0.442460</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>0.658744</td>\n",
              "      <td>0.622888</td>\n",
              "      <td>0.517898</td>\n",
              "      <td>0.405486</td>\n",
              "      <td>0.507084</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>0.656318</td>\n",
              "      <td>0.628724</td>\n",
              "      <td>0.518660</td>\n",
              "      <td>0.398013</td>\n",
              "      <td>0.506841</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.652718</td>\n",
              "      <td>0.623294</td>\n",
              "      <td>0.495050</td>\n",
              "      <td>0.395257</td>\n",
              "      <td>0.485618</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.649660</td>\n",
              "      <td>0.622728</td>\n",
              "      <td>0.507997</td>\n",
              "      <td>0.407826</td>\n",
              "      <td>0.498577</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.646614</td>\n",
              "      <td>0.611556</td>\n",
              "      <td>0.405179</td>\n",
              "      <td>0.381636</td>\n",
              "      <td>0.403265</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.643621</td>\n",
              "      <td>0.610707</td>\n",
              "      <td>0.486672</td>\n",
              "      <td>0.412258</td>\n",
              "      <td>0.480007</td>\n",
              "      <td>00:09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.640071</td>\n",
              "      <td>0.601276</td>\n",
              "      <td>0.428789</td>\n",
              "      <td>0.400854</td>\n",
              "      <td>0.426503</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.638080</td>\n",
              "      <td>0.600591</td>\n",
              "      <td>0.434882</td>\n",
              "      <td>0.403534</td>\n",
              "      <td>0.432299</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.630936</td>\n",
              "      <td>0.600719</td>\n",
              "      <td>0.450876</td>\n",
              "      <td>0.409122</td>\n",
              "      <td>0.447364</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.636138</td>\n",
              "      <td>0.593630</td>\n",
              "      <td>0.352628</td>\n",
              "      <td>0.375812</td>\n",
              "      <td>0.354309</td>\n",
              "      <td>00:09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.633998</td>\n",
              "      <td>0.600609</td>\n",
              "      <td>0.436405</td>\n",
              "      <td>0.413420</td>\n",
              "      <td>0.434547</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.631067</td>\n",
              "      <td>0.593791</td>\n",
              "      <td>0.391470</td>\n",
              "      <td>0.397525</td>\n",
              "      <td>0.391929</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.629320</td>\n",
              "      <td>0.588922</td>\n",
              "      <td>0.378522</td>\n",
              "      <td>0.404394</td>\n",
              "      <td>0.380394</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.629501</td>\n",
              "      <td>0.583223</td>\n",
              "      <td>0.336634</td>\n",
              "      <td>0.376010</td>\n",
              "      <td>0.339367</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.623397</td>\n",
              "      <td>0.576011</td>\n",
              "      <td>0.324448</td>\n",
              "      <td>0.379848</td>\n",
              "      <td>0.328129</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>0.620110</td>\n",
              "      <td>0.583899</td>\n",
              "      <td>0.389947</td>\n",
              "      <td>0.410421</td>\n",
              "      <td>0.391449</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>0.620507</td>\n",
              "      <td>0.590155</td>\n",
              "      <td>0.392993</td>\n",
              "      <td>0.409037</td>\n",
              "      <td>0.394182</td>\n",
              "      <td>00:09</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>0.620829</td>\n",
              "      <td>0.585060</td>\n",
              "      <td>0.418888</td>\n",
              "      <td>0.424710</td>\n",
              "      <td>0.419330</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>0.614943</td>\n",
              "      <td>0.577359</td>\n",
              "      <td>0.382331</td>\n",
              "      <td>0.418857</td>\n",
              "      <td>0.384913</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>0.608173</td>\n",
              "      <td>0.572798</td>\n",
              "      <td>0.367860</td>\n",
              "      <td>0.418544</td>\n",
              "      <td>0.371319</td>\n",
              "      <td>00:08</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Plot Loss\n",
        "learn.recorder.plot_losses()\n",
        "plt.show()"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 279
        },
        "id": "X5rBJVl__atj",
        "outputId": "7dbab5bc-45b2-45a2-bb53-b91007563d7f"
      },
      "execution_count": 5,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate for validation dataset\n",
        "preds_valid, _ = learn.get_preds(ds_type=DatasetType.Valid)\n",
        "pred_valid = preds_valid.argmax(dim=-1)\n",
        "\n",
        "# Evaluate Performance\n",
        "interp = ClassificationInterpretation.from_learner(learn)\n",
        "interp.plot_confusion_matrix()\n",
        "plt.show()\n",
        "\n",
        "interp.plot_tab_top_losses(5)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 564
        },
        "id": "5ITRLpaV_hip",
        "outputId": "c2906a35-3e57-4792-c054-db31a93124a2"
      },
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {}
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "  <div id=\"df-a4646c59-5272-41d6-a1d2-6ffe4af5cb77\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <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 tr th {\n",
              "        text-align: left;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr>\n",
              "      <th></th>\n",
              "      <th>Prediction</th>\n",
              "      <th>Actual</th>\n",
              "      <th>Loss</th>\n",
              "      <th>Probability</th>\n",
              "      <th>sex</th>\n",
              "      <th>education</th>\n",
              "      <th>marriage</th>\n",
              "      <th>payment_status_sep</th>\n",
              "      <th>payment_status_aug</th>\n",
              "      <th>payment_status_jul</th>\n",
              "      <th>payment_status_jun</th>\n",
              "      <th>payment_status_may</th>\n",
              "      <th>payment_status_apr</th>\n",
              "      <th>age_na</th>\n",
              "      <th>limit_bal</th>\n",
              "      <th>age</th>\n",
              "      <th>bill_statement_sep</th>\n",
              "      <th>bill_statement_aug</th>\n",
              "      <th>bill_statement_jul</th>\n",
              "      <th>bill_statement_jun</th>\n",
              "      <th>bill_statement_may</th>\n",
              "      <th>bill_statement_apr</th>\n",
              "      <th>previous_payment_sep</th>\n",
              "      <th>previous_payment_aug</th>\n",
              "      <th>previous_payment_jul</th>\n",
              "      <th>previous_payment_jun</th>\n",
              "      <th>previous_payment_may</th>\n",
              "      <th>previous_payment_apr</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>3.88</td>\n",
              "      <td>0.02</td>\n",
              "      <td>Male</td>\n",
              "      <td>school</td>\n",
              "      <td>Single</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>False</td>\n",
              "      <td>0.8693</td>\n",
              "      <td>-0.2704</td>\n",
              "      <td>-0.6816</td>\n",
              "      <td>-0.6770</td>\n",
              "      <td>-0.6704</td>\n",
              "      <td>-0.6573</td>\n",
              "      <td>-0.6568</td>\n",
              "      <td>-6.3825</td>\n",
              "      <td>-0.2794</td>\n",
              "      <td>-0.2321</td>\n",
              "      <td>-0.2390</td>\n",
              "      <td>-0.2794</td>\n",
              "      <td>-0.3085</td>\n",
              "      <td>19.4823</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>2.50</td>\n",
              "      <td>0.08</td>\n",
              "      <td>Male</td>\n",
              "      <td>school</td>\n",
              "      <td>Married</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>duly</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>False</td>\n",
              "      <td>3.2587</td>\n",
              "      <td>2.9862</td>\n",
              "      <td>7.8954</td>\n",
              "      <td>8.4296</td>\n",
              "      <td>9.3319</td>\n",
              "      <td>4.3938</td>\n",
              "      <td>5.2547</td>\n",
              "      <td>5.5822</td>\n",
              "      <td>1.3070</td>\n",
              "      <td>2.2712</td>\n",
              "      <td>17.5272</td>\n",
              "      <td>2.1539</td>\n",
              "      <td>0.9719</td>\n",
              "      <td>2.6257</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>1.59</td>\n",
              "      <td>0.20</td>\n",
              "      <td>Female</td>\n",
              "      <td>school</td>\n",
              "      <td>Married</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>False</td>\n",
              "      <td>1.7942</td>\n",
              "      <td>-0.8132</td>\n",
              "      <td>1.7745</td>\n",
              "      <td>0.5322</td>\n",
              "      <td>0.2230</td>\n",
              "      <td>-0.0783</td>\n",
              "      <td>-0.6597</td>\n",
              "      <td>-0.6506</td>\n",
              "      <td>-0.0978</td>\n",
              "      <td>-0.1631</td>\n",
              "      <td>-0.2334</td>\n",
              "      <td>-0.2902</td>\n",
              "      <td>-0.2962</td>\n",
              "      <td>-0.0842</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "      <td>1.58</td>\n",
              "      <td>0.21</td>\n",
              "      <td>Female</td>\n",
              "      <td>school</td>\n",
              "      <td>Married</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>duly</td>\n",
              "      <td>False</td>\n",
              "      <td>1.4088</td>\n",
              "      <td>-0.2704</td>\n",
              "      <td>0.4365</td>\n",
              "      <td>0.2766</td>\n",
              "      <td>-0.4162</td>\n",
              "      <td>-0.3909</td>\n",
              "      <td>-0.1552</td>\n",
              "      <td>-0.1295</td>\n",
              "      <td>3.7033</td>\n",
              "      <td>0.4965</td>\n",
              "      <td>0.7016</td>\n",
              "      <td>1.5956</td>\n",
              "      <td>1.6918</td>\n",
              "      <td>4.7714</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>1</td>\n",
              "      <td>0</td>\n",
              "      <td>1.55</td>\n",
              "      <td>0.21</td>\n",
              "      <td>Male</td>\n",
              "      <td>school</td>\n",
              "      <td>Married</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>Unknown</td>\n",
              "      <td>False</td>\n",
              "      <td>1.9484</td>\n",
              "      <td>-0.1618</td>\n",
              "      <td>5.4967</td>\n",
              "      <td>5.8922</td>\n",
              "      <td>6.3022</td>\n",
              "      <td>6.8174</td>\n",
              "      <td>7.4057</td>\n",
              "      <td>7.5982</td>\n",
              "      <td>0.8387</td>\n",
              "      <td>0.7428</td>\n",
              "      <td>0.3147</td>\n",
              "      <td>0.3140</td>\n",
              "      <td>-0.3085</td>\n",
              "      <td>0.3345</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-a4646c59-5272-41d6-a1d2-6ffe4af5cb77')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-a4646c59-5272-41d6-a1d2-6ffe4af5cb77 button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-a4646c59-5272-41d6-a1d2-6ffe4af5cb77');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ],
            "text/plain": [
              "  Prediction Actual  ... previous_payment_may previous_payment_apr\n",
              "0          0      1  ...              -0.3085              19.4823\n",
              "1          1      0  ...               0.9719               2.6257\n",
              "2          0      1  ...              -0.2962              -0.0842\n",
              "3          0      1  ...               1.6918               4.7714\n",
              "4          1      0  ...              -0.3085               0.3345\n",
              "\n",
              "[5 rows x 28 columns]"
            ]
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# Time Series Forecasting"
      ],
      "metadata": {
        "id": "Bmw3r5gT_zvw"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Data Preparation"
      ],
      "metadata": {
        "id": "Z7_e65gMkpAW"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "!pip install yfinance"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "A9IF3l6PgI2L",
        "outputId": "2cc48e60-5d87-45e0-86a3-dd49ada67dcd"
      },
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Requirement already satisfied: yfinance in /usr/local/lib/python3.7/dist-packages (0.1.68)\n",
            "Requirement already satisfied: multitasking>=0.0.7 in /usr/local/lib/python3.7/dist-packages (from yfinance) (0.0.10)\n",
            "Requirement already satisfied: lxml>=4.5.1 in /usr/local/lib/python3.7/dist-packages (from yfinance) (4.7.1)\n",
            "Requirement already satisfied: pandas>=0.24 in /usr/local/lib/python3.7/dist-packages (from yfinance) (1.1.5)\n",
            "Requirement already satisfied: numpy>=1.15 in /usr/local/lib/python3.7/dist-packages (from yfinance) (1.19.5)\n",
            "Requirement already satisfied: requests>=2.26 in /usr/local/lib/python3.7/dist-packages (from yfinance) (2.26.0)\n",
            "Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.24->yfinance) (2.8.2)\n",
            "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.24->yfinance) (2018.9)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.7/dist-packages (from python-dateutil>=2.7.3->pandas>=0.24->yfinance) (1.15.0)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.26->yfinance) (2.10)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.26->yfinance) (1.24.3)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.7/dist-packages (from requests>=2.26->yfinance) (2.0.8)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.26->yfinance) (2021.10.8)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Load Data\n",
        "\n",
        "import yfinance as yf\n",
        "import numpy as np\n",
        "\n",
        "import torch\n",
        "import torch.optim as optim\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "from torch.utils.data import (Dataset, TensorDataset, \n",
        "                              DataLoader, Subset)\n",
        "\n",
        "from sklearn.metrics import mean_squared_error\n",
        "\n",
        "print(torch.__version__)\n",
        "\n",
        "device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
        "\n",
        "# data\n",
        "TICKER = 'ANF'\n",
        "START_DATE = '2010-01-02'\n",
        "END_DATE = '2019-12-31'\n",
        "N_LAGS = 12\n",
        "\n",
        "# neural network \n",
        "VALID_SIZE = 12\n",
        "BATCH_SIZE = 5\n",
        "N_EPOCHS = 200 #1000\n",
        "\n",
        "df = yf.download(TICKER, \n",
        "                 start=START_DATE, \n",
        "                 end=END_DATE,\n",
        "                 progress=False)\n",
        "\n",
        "df = df.resample('M').last()\n",
        "prices = df['Adj Close'].values\n",
        "\n",
        "plt.plot(prices)"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 300
        },
        "id": "r1q2z4A1_5u1",
        "outputId": "0baeda55-2ac8-402a-cafe-deb58b95d494"
      },
      "execution_count": 18,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "1.10.0+cu111\n"
          ]
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[<matplotlib.lines.Line2D at 0x7f74a5e2c690>]"
            ]
          },
          "metadata": {},
          "execution_count": 18
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Convert Data to Tensor\n",
        "\n",
        "def create_input_data(series, n_lags=1):\n",
        "    '''\n",
        "    Function for transforming time series into input acceptable by a multilayer perceptron.\n",
        "    \n",
        "    Parameters\n",
        "    ----------\n",
        "    series : np.array\n",
        "        The time series to be transformed\n",
        "    n_lags : int\n",
        "        The number of lagged observations to consider as features\n",
        "\n",
        "    Returns\n",
        "    -------\n",
        "    X : np.array\n",
        "        Array of features\n",
        "    y : np.array\n",
        "        Array of target\n",
        "    '''\n",
        "    X, y = [], []\n",
        "\n",
        "    for step in range(len(series) - n_lags):\n",
        "        end_step = step + n_lags\n",
        "        X.append(series[step:end_step])\n",
        "        y.append(series[end_step])\n",
        "    return np.array(X), np.array(y)\n",
        "\n",
        "X, y = create_input_data(prices, N_LAGS)\n",
        "\n",
        "X_tensor = torch.from_numpy(X).float()\n",
        "y_tensor = torch.from_numpy(y).float().unsqueeze(dim=1)\n",
        "\n",
        "# Create Loader\n",
        "valid_ind = len(X) - VALID_SIZE\n",
        "\n",
        "dataset = TensorDataset(X_tensor, y_tensor)\n",
        "\n",
        "train_dataset = Subset(dataset, list(range(valid_ind)))\n",
        "valid_dataset = Subset(dataset, list(range(valid_ind, len(X))))\n",
        "\n",
        "train_loader = DataLoader(dataset=train_dataset,\n",
        "                          batch_size=BATCH_SIZE)\n",
        "valid_loader = DataLoader(dataset=valid_dataset, \n",
        "                          batch_size=BATCH_SIZE) "
      ],
      "metadata": {
        "id": "LoDtSkSafIDX"
      },
      "execution_count": 19,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Naive Prediction"
      ],
      "metadata": {
        "id": "hjq-ad6NgSsU"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Benchmark Naive Prediction\n",
        "naive_pred = prices[len(prices) - VALID_SIZE - 1:-1]\n",
        "y_valid = prices[len(prices) - VALID_SIZE:]\n",
        "\n",
        "naive_mse = mean_squared_error(y_valid, naive_pred)\n",
        "naive_rmse = np.sqrt(naive_mse)\n",
        "print(f\"Naive forecast - MSE: {naive_mse:.2f}, RMSE: {naive_rmse:.2f}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "wpPkvi7kfSt_",
        "outputId": "db78f4da-e407-4945-9870-f16419de668f"
      },
      "execution_count": 20,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Naive forecast - MSE: 17.37, RMSE: 4.17\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## Multi-Layer Perceptron"
      ],
      "metadata": {
        "id": "oDaGjC-BgUwS"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Define Neural Network\n",
        "class MLP(nn.Module):\n",
        "   \n",
        "    def __init__(self, input_size):\n",
        "        super(MLP, self).__init__()\n",
        "        self.linear1 = nn.Linear(input_size, 12)\n",
        "        self.linear2 = nn.Linear(12, 4)\n",
        "        self.linear3 = nn.Linear(4, 1)\n",
        "        self.dropout = nn.Dropout(p=0.2)\n",
        "    \n",
        "    def forward(self, x):\n",
        "        x = self.linear1(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.dropout(x)\n",
        "        x = self.linear2(x)\n",
        "        x = F.relu(x)\n",
        "        x = self.dropout(x)\n",
        "        x = self.linear3(x)\n",
        "        return x\n",
        "\n",
        "# set seed for reproducibility\n",
        "torch.manual_seed(42)\n",
        "\n",
        "model = MLP(N_LAGS).to(device) \n",
        "loss_fn = nn.MSELoss()\n",
        "optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
        "\n",
        "# Train\n",
        "PRINT_EVERY = 50\n",
        "train_losses, valid_losses = [], []\n",
        "\n",
        "for epoch in range(N_EPOCHS):\n",
        "    running_loss_train = 0\n",
        "    running_loss_valid = 0\n",
        "\n",
        "    model.train()\n",
        "    \n",
        "    for x_batch, y_batch in train_loader:\n",
        "        \n",
        "        optimizer.zero_grad()\n",
        "        \n",
        "        x_batch = x_batch.to(device)\n",
        "        y_batch = y_batch.to(device)\n",
        "        y_hat = model(x_batch)\n",
        "        loss = loss_fn(y_batch, y_hat)\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        running_loss_train += loss.item() * x_batch.size(0)\n",
        "        \n",
        "    epoch_loss_train = running_loss_train / len(train_loader.dataset)\n",
        "    train_losses.append(epoch_loss_train)\n",
        "\n",
        "    with torch.no_grad():\n",
        "        \n",
        "        model.eval()\n",
        "        \n",
        "        for x_val, y_val in valid_loader:\n",
        "            x_val = x_val.to(device)\n",
        "            y_val = y_val.to(device)\n",
        "            y_hat = model(x_val)\n",
        "            loss = loss_fn(y_val, y_hat)\n",
        "            running_loss_valid += loss.item() * x_val.size(0)\n",
        "            \n",
        "        epoch_loss_valid = running_loss_valid / len(valid_loader.dataset)\n",
        "            \n",
        "        if epoch > 0 and epoch_loss_valid < min(valid_losses):\n",
        "            best_epoch = epoch\n",
        "            torch.save(model.state_dict(), './mlp_checkpoint.pth')\n",
        "            \n",
        "        valid_losses.append(epoch_loss_valid)\n",
        "\n",
        "    if epoch % PRINT_EVERY == 0:\n",
        "        print(f\"<{epoch}> - Train. loss: {epoch_loss_train:.2f} \\t Valid. loss: {epoch_loss_valid:.2f}\")\n",
        "        \n",
        "print(f'Lowest loss recorded in epoch: {best_epoch}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "30OCtPyhfXUB",
        "outputId": "63b22493-91a8-42d7-b127-015e4b477cfa"
      },
      "execution_count": 21,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "<0> - Train. loss: 702.73 \t Valid. loss: 283.57\n",
            "<50> - Train. loss: 107.49 \t Valid. loss: 21.98\n",
            "<100> - Train. loss: 112.30 \t Valid. loss: 29.39\n",
            "<150> - Train. loss: 125.10 \t Valid. loss: 23.85\n",
            "Lowest loss recorded in epoch: 9\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate\n",
        "state_dict = torch.load('mlp_checkpoint.pth')\n",
        "model.load_state_dict(state_dict)\n",
        "\n",
        "y_pred, y_valid= [], []\n",
        "\n",
        "with torch.no_grad():\n",
        "\n",
        "    model.eval()\n",
        "    \n",
        "    for x_val, y_val in valid_loader:\n",
        "        x_val = x_val.to(device)    \n",
        "        y_pred.append(model(x_val))\n",
        "        y_valid.append(y_val)\n",
        "        \n",
        "y_pred = torch.cat(y_pred).numpy().flatten()\n",
        "y_valid = torch.cat(y_valid).numpy().flatten()\n",
        "\n",
        "mlp_mse = mean_squared_error(y_valid, y_pred)\n",
        "mlp_rmse = np.sqrt(mlp_mse)\n",
        "print(f\"MLP's forecast - MSE: {mlp_mse:.2f}, RMSE: {mlp_rmse:.2f}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "2gFmTmaN_53E",
        "outputId": "8be33317-6c4f-411a-f56d-f355fecea09a"
      },
      "execution_count": 22,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "MLP's forecast - MSE: 18.67, RMSE: 4.32\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## CNN"
      ],
      "metadata": {
        "id": "tfy3UcHJ_8f7"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Define CNN\n",
        "\n",
        "class Flatten(nn.Module):\n",
        "    def forward(self, x):\n",
        "        return x.view(x.size()[0], -1)\n",
        "\n",
        "model = nn.Sequential(OrderedDict([\n",
        "    ('conv_1', nn.Conv1d(1, 32, 3, padding=1)),\n",
        "    ('max_pool_1', nn.MaxPool1d(2)),\n",
        "    ('relu_1', nn.ReLU()),\n",
        "    ('flatten', Flatten()),\n",
        "    ('fc_1', nn.Linear(192, 50)),\n",
        "    ('relu_2', nn.ReLU()),\n",
        "    ('dropout_1', nn.Dropout(0.4)),\n",
        "    ('fc_2', nn.Linear(50, 1))\n",
        "]))\n",
        "\n",
        "print(model)\n",
        "\n",
        "model = model.to(device)\n",
        "loss_fn = nn.MSELoss()\n",
        "optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
        "\n",
        "# Train CNN\n",
        "PRINT_EVERY = 50\n",
        "train_losses, valid_losses = [], []\n",
        "\n",
        "for epoch in range(N_EPOCHS):\n",
        "    running_loss_train = 0\n",
        "    running_loss_valid = 0\n",
        "\n",
        "    model.train()\n",
        "    \n",
        "    for x_batch, y_batch in train_loader:\n",
        "        \n",
        "        optimizer.zero_grad()\n",
        "        \n",
        "        x_batch = x_batch.to(device)\n",
        "        x_batch = x_batch.view(x_batch.shape[0], 1, N_LAGS)\n",
        "        y_batch = y_batch.to(device)\n",
        "        y_batch = y_batch.view(y_batch.shape[0], 1, 1)\n",
        "        y_hat = model(x_batch).view(y_batch.shape[0], 1, 1)\n",
        "        loss = torch.sqrt(loss_fn(y_batch, y_hat))\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        running_loss_train += loss.item() * x_batch.size(0)\n",
        "        \n",
        "    epoch_loss_train = running_loss_train / len(train_loader.dataset)\n",
        "    train_losses.append(epoch_loss_train)\n",
        "\n",
        "    with torch.no_grad():\n",
        "        model.eval()\n",
        "        for x_val, y_val in valid_loader:\n",
        "            x_val = x_val.to(device)\n",
        "            x_val = x_val.view(x_val.shape[0], 1, N_LAGS)\n",
        "            y_val = y_val.to(device)\n",
        "            y_val = y_val.view(y_val.shape[0], 1, 1)\n",
        "            y_hat = model(x_val).view(y_val.shape[0], 1, 1)\n",
        "            loss = torch.sqrt(loss_fn(y_val, y_hat))\n",
        "            running_loss_valid += loss.item() * x_val.size(0)\n",
        "            \n",
        "        epoch_loss_valid = running_loss_valid / len(valid_loader.dataset)\n",
        "            \n",
        "        if epoch > 0 and epoch_loss_valid < min(valid_losses):\n",
        "            best_epoch = epoch\n",
        "            torch.save(model.state_dict(), './cnn_checkpoint.pth')\n",
        "            \n",
        "        valid_losses.append(epoch_loss_valid)\n",
        "\n",
        "    if epoch % PRINT_EVERY == 0:\n",
        "        print(f\"<{epoch}> - Train. loss: {epoch_loss_train:.6f} \\t Valid. loss: {epoch_loss_valid:.6f}\")\n",
        "        \n",
        "print(f'Lowest loss recorded in epoch: {best_epoch}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "LWRu_pXY_59f",
        "outputId": "1b962b97-564b-45b3-f110-f2e448892bfe"
      },
      "execution_count": 23,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Sequential(\n",
            "  (conv_1): Conv1d(1, 32, kernel_size=(3,), stride=(1,), padding=(1,))\n",
            "  (max_pool_1): MaxPool1d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n",
            "  (relu_1): ReLU()\n",
            "  (flatten): Flatten()\n",
            "  (fc_1): Linear(in_features=192, out_features=50, bias=True)\n",
            "  (relu_2): ReLU()\n",
            "  (dropout_1): Dropout(p=0.4, inplace=False)\n",
            "  (fc_2): Linear(in_features=50, out_features=1, bias=True)\n",
            ")\n",
            "<0> - Train. loss: 14.521913 \t Valid. loss: 6.024386\n",
            "<50> - Train. loss: 5.891373 \t Valid. loss: 4.508741\n",
            "<100> - Train. loss: 6.438412 \t Valid. loss: 4.286388\n",
            "<150> - Train. loss: 6.304438 \t Valid. loss: 4.349677\n",
            "Lowest loss recorded in epoch: 2\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate\n",
        "state_dict = torch.load('cnn_checkpoint.pth')\n",
        "model.load_state_dict(state_dict)\n",
        "\n",
        "y_pred, y_valid = [], []\n",
        "\n",
        "with torch.no_grad():\n",
        "    \n",
        "    model.eval()\n",
        "        \n",
        "    for x_val, y_val in valid_loader:\n",
        "        x_val = x_val.to(device)\n",
        "        x_val = x_val.view(x_val.shape[0], 1, N_LAGS)\n",
        "        y_pred.append(model(x_val))\n",
        "        y_valid.append(y_val)\n",
        "        \n",
        "y_pred = torch.cat(y_pred).numpy().flatten()\n",
        "y_valid = torch.cat(y_valid).numpy().flatten()\n",
        "\n",
        "mlp_mse = mean_squared_error(y_valid, y_pred)\n",
        "mlp_rmse = np.sqrt(mlp_mse)\n",
        "print(f\"CNN's forecast - MSE: {mlp_mse:.2f}, RMSE: {mlp_rmse:.2f}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "r6tsZtWjAEaL",
        "outputId": "565707da-ef6a-4a32-e518-f8a64725f060"
      },
      "execution_count": 24,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CNN's forecast - MSE: 19.14, RMSE: 4.37\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "## RNN"
      ],
      "metadata": {
        "id": "LFvGAp9z_9KP"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# Create Model\n",
        "class RNN(nn.Module):\n",
        "    def __init__(self, input_size, hidden_size, n_layers, output_size):\n",
        "        super(RNN, self).__init__()\n",
        "        self.rnn = nn.RNN(input_size, hidden_size, \n",
        "                          n_layers, batch_first=True,\n",
        "                          nonlinearity='relu')\n",
        "        self.fc = nn.Linear(hidden_size, output_size)\n",
        "    \n",
        "    def forward(self, x):\n",
        "        output, _ = self.rnn(x)\n",
        "        output = self.fc(output[:,-1,:]) \n",
        "        return output\n",
        "\n",
        "model = RNN(input_size=1, hidden_size=6, \n",
        "            n_layers=1, output_size=1).to(device)\n",
        "loss_fn = nn.MSELoss()\n",
        "optimizer = optim.Adam(model.parameters(), lr=0.001)\n",
        "\n",
        "# Train\n",
        "PRINT_EVERY = 10\n",
        "train_losses, valid_losses = [], []\n",
        "\n",
        "for epoch in range(N_EPOCHS):\n",
        "    running_loss_train = 0\n",
        "    running_loss_valid = 0\n",
        "\n",
        "    model.train()\n",
        "    \n",
        "    for x_batch, y_batch in train_loader:\n",
        "        \n",
        "        optimizer.zero_grad()\n",
        "        \n",
        "        x_batch = x_batch.to(device)[:, :, None]\n",
        "        y_batch = y_batch.to(device)\n",
        "        y_hat = model(x_batch)\n",
        "        loss = torch.sqrt(loss_fn(y_batch, y_hat))\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "        running_loss_train += loss.item() * x_batch.size(0)\n",
        "        \n",
        "    epoch_loss_train = running_loss_train / len(train_loader.dataset)\n",
        "    train_losses.append(epoch_loss_train)\n",
        "\n",
        "    with torch.no_grad():\n",
        "        model.eval()\n",
        "        for x_val, y_val in valid_loader:\n",
        "            x_val = x_val.to(device)[:, :, None]\n",
        "            y_val = y_val.to(device)\n",
        "            y_hat = model(x_val)\n",
        "            loss = torch.sqrt(loss_fn(y_val, y_hat))\n",
        "            running_loss_valid += loss.item() * x_val.size(0)\n",
        "            \n",
        "        epoch_loss_valid = running_loss_valid / len(valid_loader.dataset)\n",
        "            \n",
        "        if epoch > 0 and epoch_loss_valid < min(valid_losses):\n",
        "            best_epoch = epoch\n",
        "            torch.save(model.state_dict(), './rnn_checkpoint.pth')\n",
        "            \n",
        "        valid_losses.append(epoch_loss_valid)\n",
        "\n",
        "    if epoch % PRINT_EVERY == 0:\n",
        "        print(f\"<{epoch}> - Train. loss: {epoch_loss_train:.4f} \\t Valid. loss: {epoch_loss_valid:.4f}\")\n",
        "        \n",
        "print(f'Lowest loss recorded in epoch: {best_epoch}')"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lXmDKTtBAD6L",
        "outputId": "d96ae86f-608e-4d54-acd7-9e086f80614d"
      },
      "execution_count": 25,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "<0> - Train. loss: 25.5478 \t Valid. loss: 18.3735\n",
            "<10> - Train. loss: 25.3490 \t Valid. loss: 18.1754\n",
            "<20> - Train. loss: 25.1503 \t Valid. loss: 17.9774\n",
            "<30> - Train. loss: 24.9517 \t Valid. loss: 17.7794\n",
            "<40> - Train. loss: 24.7531 \t Valid. loss: 17.5814\n",
            "<50> - Train. loss: 24.5545 \t Valid. loss: 17.3835\n",
            "<60> - Train. loss: 24.3560 \t Valid. loss: 17.1857\n",
            "<70> - Train. loss: 24.1575 \t Valid. loss: 16.9879\n",
            "<80> - Train. loss: 23.9590 \t Valid. loss: 16.7902\n",
            "<90> - Train. loss: 23.7606 \t Valid. loss: 16.5925\n",
            "<100> - Train. loss: 23.5622 \t Valid. loss: 16.3949\n",
            "<110> - Train. loss: 23.3639 \t Valid. loss: 16.1974\n",
            "<120> - Train. loss: 23.1656 \t Valid. loss: 15.9999\n",
            "<130> - Train. loss: 22.9673 \t Valid. loss: 15.8024\n",
            "<140> - Train. loss: 22.7691 \t Valid. loss: 15.6051\n",
            "<150> - Train. loss: 22.5710 \t Valid. loss: 15.4077\n",
            "<160> - Train. loss: 22.3728 \t Valid. loss: 15.2105\n",
            "<170> - Train. loss: 22.1748 \t Valid. loss: 15.0133\n",
            "<180> - Train. loss: 21.9768 \t Valid. loss: 14.8162\n",
            "<190> - Train. loss: 21.7789 \t Valid. loss: 14.6192\n",
            "Lowest loss recorded in epoch: 199\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# Evaluate\n",
        "state_dict = torch.load('rnn_checkpoint.pth')\n",
        "model.load_state_dict(state_dict)\n",
        "\n",
        "y_pred = []\n",
        "\n",
        "with torch.no_grad():\n",
        "    \n",
        "    model.eval()\n",
        "    \n",
        "    for x_val, y_val in valid_loader:\n",
        "        x_val = x_val.to(device)[:, :, None]\n",
        "        y_hat = model(x_val)\n",
        "        y_pred.append(y_hat)\n",
        "        \n",
        "y_pred = torch.cat(y_pred).numpy().flatten()\n",
        "# y_pred = minmax.inverse_transform(y_pred).flatten()\n",
        "\n",
        "mlp_mse = mean_squared_error(y_valid, y_pred)\n",
        "mlp_rmse = np.sqrt(mlp_mse)\n",
        "print(f\"CNN's forecast - MSE: {mlp_mse:.2f}, RMSE: {mlp_rmse:.2f}\")"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VwDm0-c-_wKo",
        "outputId": "7c32b82b-a0dd-4100-c38b-acb78c36c08a"
      },
      "execution_count": 26,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CNN's forecast - MSE: 220.35, RMSE: 14.84\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        ""
      ],
      "metadata": {
        "id": "yyEXlHSxjAH9"
      },
      "execution_count": 26,
      "outputs": []
    }
  ]
}