misc/feature-selection/feature_selection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "b036c95c-d086-4e51-90c3-bc24a256de9e",
   "metadata": {},
   "source": [
    "# Feature Reduction\n",
    "\n",
    "This notebook demonstrates how to use various feature selection methods using scalecast. The idea behind this example is to reduce the variables stored in the `Forecaster` object to an optimal subset after adding many potentially helpful features to the object.  \n",
    "\n",
    "- Link to [data](https://www.kaggle.com/robervalt/sunspots)    \n",
    "- See the [blog post](https://medium.com/towards-data-science/variable-reduction-techniques-for-time-series-646743f726d4)  \n",
    "- [Function documentation link](https://scalecast.readthedocs.io/en/latest/Forecaster/Forecaster.html#src.scalecast.Forecaster.Forecaster.reduce_Xvars)  \n",
    "\n",
    "![](mermaid-diagram-2023-02-05-132837.png)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "c6d85dc9-8631-41a3-9e6a-466b1b60752b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from scalecast.Forecaster import Forecaster\n",
    "from scalecast.util import plot_reduction_errors"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "b05f1eaf-fa70-4eab-8a65-5fbe93d4065c",
   "metadata": {},
   "outputs": [],
   "source": [
    "sns.set(rc={\"figure.figsize\": (12, 8)})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "be48f350-4d2e-499b-9a4d-f4d24b674a0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "def prepare_fcst(f, test_length=120, fcst_length=120, validation_length=120):\n",
    "    \"\"\" adds all variables and sets the test length/forecast length in the object\n",
    "    \n",
    "    Args:\n",
    "        f (Forecaster): the Forecaster object.\n",
    "        test_length (int or float): the test length as a size or proportion.\n",
    "        fcst_length (int): the forecast horizon.\n",
    "        \n",
    "    Returns:\n",
    "        (Forecaster) the processed object.\n",
    "    \"\"\"\n",
    "    f.generate_future_dates(fcst_length)\n",
    "    f.set_test_length(test_length)\n",
    "    f.set_validation_length(validation_length)\n",
    "    f.add_seasonal_regressors(\"month\", \"quarter\", raw=False, sincos=True)\n",
    "    f.set_validation_metric(\"mae\")\n",
    "    for i in np.arange(12, 289, 12):  # 12-month cycles from 12 to 288 months\n",
    "        f.add_cycle(i)\n",
    "    f.add_ar_terms(120)  # AR 1-120\n",
    "    f.add_AR_terms((20, 12))  # seasonal AR up to 20 years, spaced one year apart\n",
    "    f.add_time_trend()\n",
    "    f.add_seasonal_regressors(\"year\")\n",
    "    return f\n",
    "\n",
    "\n",
    "def export_results(f):\n",
    "    \"\"\" returns a dataframe with all model results given a Forecaster object.\n",
    "    \n",
    "    Args:\n",
    "        f (Forecaster): the Forecaster object.\n",
    "        \n",
    "    Returns:\n",
    "        (DataFrame) the dataframe with the pertinent results.\n",
    "    \"\"\"\n",
    "    results = f.export(\"model_summaries\", determine_best_by=\"TestSetMAE\")\n",
    "    results[\"N_Xvars\"] = results[\"Xvars\"].apply(lambda x: len(x))\n",
    "    return results[\n",
    "        [\n",
    "            \"ModelNickname\",\n",
    "            \"TestSetMAE\",\n",
    "            \"InSampleMAE\",\n",
    "            \"TestSetR2\",\n",
    "            \"InSampleR2\",\n",
    "            \"DynamicallyTested\",\n",
    "            \"N_Xvars\",\n",
    "        ]\n",
    "    ]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c690e6b3-4674-40fe-b070-cbfea6581331",
   "metadata": {},
   "source": [
    "## Load Forecaster Object\n",
    "- we choose 120 periods (10 years) for all validation and forecasting\n",
    "- 10 years of observervations to tune model hyperparameters, 10 years to test, and a forecast horizon of 10 years"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "cc3f237f-ea02-4aef-b237-3a103b799e31",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Forecaster(\n",
       "    DateStartActuals=1749-01-31T00:00:00.000000000\n",
       "    DateEndActuals=2021-01-31T00:00:00.000000000\n",
       "    Freq=M\n",
       "    N_actuals=3265\n",
       "    ForecastLength=120\n",
       "    Xvars=['monthsin', 'monthcos', 'quartersin', 'quartercos', 'cycle12sin', 'cycle12cos', 'cycle24sin', 'cycle24cos', 'cycle36sin', 'cycle36cos', 'cycle48sin', 'cycle48cos', 'cycle60sin', 'cycle60cos', 'cycle72sin', 'cycle72cos', 'cycle84sin', 'cycle84cos', 'cycle96sin', 'cycle96cos', 'cycle108sin', 'cycle108cos', 'cycle120sin', 'cycle120cos', 'cycle132sin', 'cycle132cos', 'cycle144sin', 'cycle144cos', 'cycle156sin', 'cycle156cos', 'cycle168sin', 'cycle168cos', 'cycle180sin', 'cycle180cos', 'cycle192sin', 'cycle192cos', 'cycle204sin', 'cycle204cos', 'cycle216sin', 'cycle216cos', 'cycle228sin', 'cycle228cos', 'cycle240sin', 'cycle240cos', 'cycle252sin', 'cycle252cos', 'cycle264sin', 'cycle264cos', 'cycle276sin', 'cycle276cos', 'cycle288sin', 'cycle288cos', 'AR1', 'AR2', 'AR3', 'AR4', 'AR5', 'AR6', 'AR7', 'AR8', 'AR9', 'AR10', 'AR11', 'AR12', 'AR13', 'AR14', 'AR15', 'AR16', 'AR17', 'AR18', 'AR19', 'AR20', 'AR21', 'AR22', 'AR23', 'AR24', 'AR25', 'AR26', 'AR27', 'AR28', 'AR29', 'AR30', 'AR31', 'AR32', 'AR33', 'AR34', 'AR35', 'AR36', 'AR37', 'AR38', 'AR39', 'AR40', 'AR41', 'AR42', 'AR43', 'AR44', 'AR45', 'AR46', 'AR47', 'AR48', 'AR49', 'AR50', 'AR51', 'AR52', 'AR53', 'AR54', 'AR55', 'AR56', 'AR57', 'AR58', 'AR59', 'AR60', 'AR61', 'AR62', 'AR63', 'AR64', 'AR65', 'AR66', 'AR67', 'AR68', 'AR69', 'AR70', 'AR71', 'AR72', 'AR73', 'AR74', 'AR75', 'AR76', 'AR77', 'AR78', 'AR79', 'AR80', 'AR81', 'AR82', 'AR83', 'AR84', 'AR85', 'AR86', 'AR87', 'AR88', 'AR89', 'AR90', 'AR91', 'AR92', 'AR93', 'AR94', 'AR95', 'AR96', 'AR97', 'AR98', 'AR99', 'AR100', 'AR101', 'AR102', 'AR103', 'AR104', 'AR105', 'AR106', 'AR107', 'AR108', 'AR109', 'AR110', 'AR111', 'AR112', 'AR113', 'AR114', 'AR115', 'AR116', 'AR117', 'AR118', 'AR119', 'AR120', 'AR132', 'AR144', 'AR156', 'AR168', 'AR180', 'AR192', 'AR204', 'AR216', 'AR228', 'AR240', 't', 'year']\n",
       "    TestLength=120\n",
       "    ValidationMetric=mae\n",
       "    ForecastsEvaluated=[]\n",
       "    CILevel=None\n",
       "    CurrentEstimator=mlr\n",
       "    GridsFile=Grids\n",
       ")"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.read_csv(\"Sunspots.csv\", index_col=0, names=[\"Date\", \"Target\"], header=0)\n",
    "f = Forecaster(y=df[\"Target\"], current_dates=df[\"Date\"])\n",
    "prepare_fcst(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "c48e5286-604d-4f18-ab9e-b3b9f039d4a3",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "starting out with 184 variables\n"
     ]
    }
   ],
   "source": [
    "print(\"starting out with {} variables\".format(len(f.get_regressor_names())))"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cac05d69-82a1-4f90-8bae-c45294f6d9cc",
   "metadata": {},
   "source": [
    "## Reducing with L1 Regularization\n",
    "- L1 regularization reduces out least important variables using a lasso model\n",
    "- Not computationally expensive\n",
    "- `overwrite=False` is used to not replace variables currently in the `Forecaster` object"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "e71da70f-0d1b-4e89-8515-2e771a221c0e",
   "metadata": {},
   "outputs": [],
   "source": [
    "f.reduce_Xvars(overwrite=False, dynamic_testing=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "767a8f81-6f3e-4e15-ab83-d7bbf301efb1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "lasso reduced to 59 variables\n"
     ]
    }
   ],
   "source": [
    "lasso_reduced_vars = f.reduced_Xvars[:]\n",
    "print(f\"lasso reduced to {len(lasso_reduced_vars)} variables\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2ff5d0c6-63ce-408b-bf6b-1b2963d47d46",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "monthcos\n",
      "quartersin\n",
      "cycle12cos\n",
      "cycle24sin\n",
      "cycle24cos\n",
      "cycle36cos\n",
      "cycle48cos\n",
      "cycle60sin\n",
      "cycle60cos\n",
      "cycle72sin\n",
      "cycle84sin\n",
      "cycle84cos\n",
      "cycle96sin\n",
      "cycle96cos\n",
      "cycle108sin\n",
      "cycle108cos\n",
      "cycle120sin\n",
      "cycle120cos\n",
      "cycle132sin\n",
      "cycle132cos\n",
      "cycle144sin\n",
      "cycle156sin\n",
      "cycle168cos\n",
      "cycle180sin\n",
      "cycle192cos\n",
      "cycle204sin\n",
      "cycle216sin\n",
      "cycle216cos\n",
      "cycle228sin\n",
      "cycle252sin\n",
      "cycle252cos\n",
      "cycle264cos\n",
      "cycle288sin\n",
      "cycle288cos\n",
      "AR1\n",
      "AR2\n",
      "AR3\n",
      "AR4\n",
      "AR5\n",
      "AR6\n",
      "AR8\n",
      "AR9\n",
      "AR10\n",
      "AR11\n",
      "AR18\n",
      "AR21\n",
      "AR27\n",
      "AR28\n",
      "AR29\n",
      "AR34\n",
      "AR35\n",
      "AR92\n",
      "AR102\n",
      "AR108\n",
      "AR111\n",
      "AR113\n",
      "AR116\n",
      "AR117\n",
      "AR216\n"
     ]
    }
   ],
   "source": [
    "print(*lasso_reduced_vars, sep=\"\\n\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c8f0933f-9412-40d6-a94d-1bd122807efb",
   "metadata": {},
   "source": [
    "## Reducing with Permutation Feature Importance\n",
    "- uses the [ELI5 package](https://github.com/eli5-org/eli5) to score feature importances\n",
    "- eliminates the least-scoring variable (with some scalecast adjustment to account for collinearity), re-trains the model, and evaluates the new error\n",
    "- by default, the Xvars that return the best error score (according to any metric you want to monitor) is chosen as the optimal set of variables\n",
    "- to save computation resources, not every variable combination is tried (unlike other reduction techniques)\n",
    "- to save computation resources, we can set `dynamic_testing = False` and we can set a mininum number of variables to use (the square root of the total number of observations is chosen in this example)\n",
    "- we can use any sklearn model to reduce variables using this method and we select MLR, KNN, GBT, and SVR in this example"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "1fe85242-0a2b-4e42-bd09-22d0102c4eb4",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2023-04-11 12:03:53.245099: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.1 SSE4.2\n",
      "To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.\n"
     ]
    }
   ],
   "source": [
    "f.reduce_Xvars(\n",
    "    method=\"pfi\",\n",
    "    estimator=\"mlr\",\n",
    "    keep_at_least=\"sqrt\",\n",
    "    grid_search=False,\n",
    "    normalizer=\"minmax\",\n",
    "    overwrite=False,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4a32e2e2-ec75-45c9-b03f-2c6e2e90a1e0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "mlr with pfi reduced to 166 variables\n"
     ]
    }
   ],
   "source": [
    "mlr_reduced_vars = f.reduced_Xvars[:]\n",
    "print(f\"mlr with pfi reduced to {len(mlr_reduced_vars)} variables\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "7a20d27a-51ec-4941-9920-f811c373e6fb",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1200x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_reduction_errors(f)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "97d571e6-5d97-4290-8138-34573f379497",
   "metadata": {},
   "source": [
    "The above graph shows that the MLR performs best when about 50 variables are reduced, according to the MAE on the validation set. The order the variables were dropped was based on permutation feature importance. Clearly, too many variables had been added initially and reducing them out made a significant positive difference to the model's performance. Note, the error metric monitored is not from the test set, but from a different out-of-sample slice of data. Later models use the test set to validate the decisions that were made while monitoring this validation metric."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "e86e4737-c0d4-4523-8097-d562622da2a3",
   "metadata": {},
   "outputs": [],
   "source": [
    "def reduce_and_save_results(f, model, results):\n",
    "    f.reduce_Xvars(\n",
    "        method=\"pfi\",\n",
    "        estimator=model,\n",
    "        keep_at_least=\"sqrt\",\n",
    "        grid_search=True,\n",
    "        dynamic_testing=False,\n",
    "        overwrite=False,\n",
    "    )\n",
    "    results[model] = [\n",
    "        f.reduced_Xvars,\n",
    "        f.pfi_dropped_vars,\n",
    "        f.pfi_error_values,\n",
    "        f.reduction_hyperparams,\n",
    "    ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "a6c8d9a5-bd3c-4475-88f4-00e36ebbf9e7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "reducing with knn\n"
     ]
    }
   ],
   "source": [
    "results = {\n",
    "    \"mlr\": [\n",
    "        f.reduced_Xvars,\n",
    "        f.pfi_dropped_vars,\n",
    "        f.pfi_error_values,\n",
    "        f.reduction_hyperparams,\n",
    "    ]\n",
    "}\n",
    "\n",
    "for model in (\"knn\", \"gbt\", \"svr\"):\n",
    "    print(f\"reducing with {model}\")\n",
    "    reduce_and_save_results(f, model, results)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "c93d9450-2ab6-4f7d-8f34-17549eb26d62",
   "metadata": {},
   "outputs": [],
   "source": [
    "for k, v in results.items():\n",
    "    print(f\"the {k} model chose {len(v[0])} variables\")\n",
    "    sns.lineplot(x=np.arange(0, len(v[1]) + 1, 1), y=v[2], label=k)\n",
    "plt.xlabel(\"dropped Xvars\")\n",
    "plt.ylabel(\"error\")\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "cf971c3b-daa8-4b10-9451-06d836de76b9",
   "metadata": {},
   "source": [
    "Out of the four models we tried using pfi, the gbt model scored the best on the validation set with 136 variables. We will re-train that model with each of the reduced set of variables and see which reduction performs best on the test set. We will use the best-performing model class from the four we tried: gbt. There is no perfect way to choose a set of variables and it is clear that the gbt model performs only marginally different depending on how many variables are dropped, probably because of how the underlying decision tree model can give lower weight to less-important features natively."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "8f87f52c-0df3-4da9-9380-8dca96910059",
   "metadata": {},
   "outputs": [],
   "source": [
    "knn_reduced_vars = results[\"knn\"][0]\n",
    "gbt_reduced_vars = results[\"gbt\"][0]\n",
    "svr_reduced_vars = results[\"svr\"][0]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e97da2df-b2a3-4559-a98a-5f96be971b63",
   "metadata": {},
   "source": [
    "## Training a Model Class with All Reduced Variable Sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "01ef31c0-4033-4a1f-b27f-b9fddffc0c44",
   "metadata": {},
   "outputs": [],
   "source": [
    "selected_model = \"gbt\"\n",
    "hp = results[selected_model][3]\n",
    "f.set_estimator(selected_model)\n",
    "f.manual_forecast(**hp, Xvars=\"all\", call_me=selected_model + \"_all_vars\")\n",
    "f.manual_forecast(\n",
    "    **hp, Xvars=lasso_reduced_vars, call_me=selected_model + \"_l1_reduced_vars\"\n",
    ")\n",
    "f.manual_forecast(\n",
    "    **hp, Xvars=mlr_reduced_vars, call_me=selected_model + \"pfi-mlr_reduced_vars\"\n",
    ")\n",
    "f.manual_forecast(\n",
    "    **hp, Xvars=knn_reduced_vars, call_me=selected_model + \"pfi-knn_reduced_vars\"\n",
    ")\n",
    "f.manual_forecast(\n",
    "    **hp, Xvars=gbt_reduced_vars, call_me=selected_model + \"pfi-gbt_reduced_vars\"\n",
    ")\n",
    "f.manual_forecast(\n",
    "    **hp, Xvars=svr_reduced_vars, call_me=selected_model + \"pfi-svr_reduced_vars\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7fe0eb3a-ce56-41ba-b26e-d0424bf114d0",
   "metadata": {},
   "outputs": [],
   "source": [
    "export_results(f)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ee0ad7d2-a5b7-4662-86a9-545830956725",
   "metadata": {},
   "source": [
    "Although the gbt model with 136 variables performed best out of all tried models on the validation set, on the test set, the 98 variables selected by KNN led to the best performance, followed by using all variables and then the 36 selected from the L1 reduction. This illustrates that even if a reduction decision is made with one estimator class, it can still be useful for another model class."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "0a909fa1-daa9-449f-a0a4-df1df7eeefc8",
   "metadata": {},
   "outputs": [],
   "source": [
    "f.plot_test_set(ci=True, order_by=\"TestSetMAE\", include_train=240)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "60646747-d4a4-45d9-a853-e25708e7fd6f",
   "metadata": {},
   "source": [
    "## Backtest Best Model\n",
    "\n",
    "Backtesting shows the average performance of a given model across the last 10 (by default) forecast horizons, using only data before each 120-period forecast to train the model. It's a good way to see if your model generalizes well or just got lucky on the particular test set you fed to it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "bd11acfd-b145-409c-a8fa-86ee56b0db1c",
   "metadata": {},
   "outputs": [],
   "source": [
    "f.backtest(\"gbtpfi-knn_reduced_vars\")\n",
    "f.export_backtest_metrics(\"gbtpfi-knn_reduced_vars\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "1f2e4368-4907-42d4-8667-a657ea44d5e1",
   "metadata": {},
   "outputs": [],
   "source": [
    "f.plot(models=\"gbtpfi-knn_reduced_vars\", ci=True)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "90e881c4-b6f4-452e-90c5-616d9b40943b",
   "metadata": {},
   "source": [
    "## Export Reduced Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "699bdaf6-e697-41c2-80b9-f984c203d532",
   "metadata": {},
   "outputs": [],
   "source": [
    "reduced_dataset = f.export_Xvars_df()[['DATE'] + knn_reduced_vars]\n",
    "reduced_dataset.head(25)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b57824f5-2f0e-46ef-98b8-cef92ad823a7",
   "metadata": {},
   "source": [
    "## See how often each var was dropped"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "28913029-9e27-4267-b899-bb9fecbe9ed4",
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.options.display.max_rows = None\n",
    "all_xvars = f.get_regressor_names()\n",
    "final_dropped = pd.DataFrame({\"Var\": all_xvars})\n",
    "for i, v in f.export(\"model_summaries\").iterrows():\n",
    "    model = v[\"ModelNickname\"]\n",
    "    Xvars = v[\"Xvars\"]\n",
    "    dropped_vars = [x for x in f.get_regressor_names() if x not in Xvars]\n",
    "    if not dropped_vars:\n",
    "        continue\n",
    "    tmp_dropped = pd.DataFrame(\n",
    "        {\"Var\": dropped_vars, f\"dropped in {model}\": [1] * len(dropped_vars)}\n",
    "    )\n",
    "    final_dropped = final_dropped.merge(tmp_dropped, on=\"Var\", how=\"left\").fillna(0)\n",
    "final_dropped[\"total times dropped\"] = final_dropped.iloc[:, 1:].sum(axis=1)\n",
    "final_dropped = final_dropped.loc[final_dropped[\"total times dropped\"] > 0]\n",
    "final_dropped = final_dropped.sort_values(\"total times dropped\", ascending=False)\n",
    "final_dropped = final_dropped.reset_index(drop=True)\n",
    "final_dropped.iloc[:, 1:] = final_dropped.iloc[:, 1:].astype(int)\n",
    "final_dropped"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "84af132c-faf9-4299-bc9b-f973358fe658",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.13"
  },
  "widgets": {
   "application/vnd.jupyter.widget-state+json": {
    "state": {},
    "version_major": 2,
    "version_minor": 0
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}