transactions feature selection.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This feature selection notebook does a filter followed by a wrapper for a binary dependent variable (binary classification). It's capable of doing the filter on more than one file. The variable files are called vars1.csv, vars2.csv ... Or you can make the input file name(s) anything you want.\n",
    "\n",
    "The filter runs separately on each vars file and keeps the top num_filter variables from each file. If there are more than one vars files we'll again select the top num_filter variables across all the vars.csv files.\n",
    "\n",
    "If balance = 0 the entire files are used. If balance != 0 then balance is the RATIO OF BADS TO GOODS retained for the rest of the feature selection. We keep all the rare class (bads) and downsample the goods. I think in general it's better to keep balance = 0.\n",
    "\n",
    "I've got an annoying warning message from the wrapper and I can't figure out how to get rid of it. If anybody figures this out please send a message to stevecoggeshall@gmail.com"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting lightgbm\n",
      "  Downloading lightgbm-3.3.5-py3-none-win_amd64.whl (1.0 MB)\n",
      "     ---------------------------------------- 1.0/1.0 MB 3.1 MB/s eta 0:00:00\n",
      "Requirement already satisfied: scipy in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from lightgbm) (1.9.1)\n",
      "Requirement already satisfied: numpy in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from lightgbm) (1.21.5)\n",
      "Requirement already satisfied: wheel in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from lightgbm) (0.37.1)\n",
      "Requirement already satisfied: scikit-learn!=0.22.0 in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from lightgbm) (1.0.2)\n",
      "Requirement already satisfied: threadpoolctl>=2.0.0 in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from scikit-learn!=0.22.0->lightgbm) (2.2.0)\n",
      "Requirement already satisfied: joblib>=0.11 in c:\\users\\rijul\\anaconda3\\lib\\site-packages (from scikit-learn!=0.22.0->lightgbm) (1.1.0)\n",
      "Installing collected packages: lightgbm\n",
      "Successfully installed lightgbm-3.3.5\n"
     ]
    }
   ],
   "source": [
    "\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import scipy.stats as sps\n",
    "import matplotlib.pyplot as plt\n",
    "import datetime as dt\n",
    "import gc\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from mlxtend.feature_selection import SequentialFeatureSelector as SFS\n",
    "from lightgbm import LGBMClassifier\n",
    "%matplotlib inline\n",
    "start_time = dt.datetime.now()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set some parameters\n",
    "num_files = 1\n",
    "# I recommend set num_filter to be about 10 to 20% of the original # variables\n",
    "num_filter = 200\n",
    "# I recommend set num_wrapper to be about 50, then look for a saturation of the model performance as variables are added\n",
    "# Then you can run it again with num_wrapper just a bit above this saturation point, not more than about twice this saturation number\n",
    "num_wrapper = 20\n",
    "balance = 0\n",
    "detect_rate = .03\n",
    "index_name = 'Recnum'\n",
    "y_name = 'Fraud'\n",
    "good_label = 0\n",
    "bad_label = 1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run a filter on all the files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%%time\n",
    "filter_score_df_list = []\n",
    "for i in range(num_files):\n",
    "#     file_name = \"vars\"+str(i+1)+'.csv'\n",
    "    file_name = 'candidate_variables.csv'\n",
    "    df = pd.read_csv(file_name)\n",
    "    print(\"********** working on\",file_name,\"size is\",df.shape)\n",
    "    df = df.set_index(index_name) \n",
    "#     df = df[df.index <= 84300] # remove the last two months as the out-of-time data (OOT)\n",
    "    df = df[df.index >= 2995] # remove the first 2 weeks of records since their variables aren't well formed\n",
    "    df['RANDOM'] = np.random.ranf(len(df)) # add a random number variable to make sure it doesn't come up as important\n",
    "    goods = df[df[y_name] == good_label]\n",
    "    bads = df[df[y_name] == bad_label]\n",
    "    del df # don't need this file anymore\n",
    "    num_goods = len(goods)\n",
    "    num_bads = len(bads)\n",
    "    num_vars = len(bads.columns)-2\n",
    "    if(balance != 0):\n",
    "        if(i == 0):\n",
    "            num_goods_desired = int(min(num_goods,num_bads*balance))\n",
    "            goods = goods.sample(n=num_goods_desired,random_state=1)\n",
    "            goods_keep = list(goods.index)\n",
    "            goods_keep.sort()\n",
    "    \n",
    "        if(i > 0):\n",
    "            goods = goods.loc[goods_keep] \n",
    "            \n",
    "    df_sampled = pd.concat([goods,bads])\n",
    "    df_sampled.sort_index(inplace=True)\n",
    "    filter_score = pd.DataFrame(np.zeros((num_vars+1,2)))\n",
    "    filter_score.columns = ['variable','filter score']   \n",
    "    j = 0\n",
    "    for column in df_sampled:\n",
    "        filter_score.loc[j,'variable'] = column\n",
    "        filter_score.loc[j,'filter score'] = sps.ks_2samp(goods[column],bads[column])[0]\n",
    "        j = j+1\n",
    "        if j%100 == 0:\n",
    "            print(j)\n",
    "\n",
    "    filter_score.sort_values(by=['filter score'], ascending=False, inplace=True)\n",
    "    vars_keep = list(filter_score['variable'][1:num_filter+1]) \n",
    "    print(file_name,filter_score.head(20))\n",
    "    if(i == 0): # if first time through need to initialize some stuff\n",
    "        Y = pd.DataFrame(df_sampled[y_name], index=df_sampled.index)\n",
    "        df_top = df_sampled.filter(vars_keep, axis=1)\n",
    "            \n",
    "    if(i > 0): # if more than one variable file we use this loop\n",
    "        data_new_top = df_sampled.filter(vars_keep, axis=1)\n",
    "        df_top = pd.concat([df_top,data_new_top], axis=1)\n",
    "\n",
    "    filter_score_df_list.append(filter_score)\n",
    "    \n",
    "    del goods # delete these before starting the next file, if any\n",
    "    del bads\n",
    "    gc.collect()\n",
    "filter_score = pd.concat(filter_score_df_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "ename": "NameError",
     "evalue": "name 'filter_score' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "\u001b[1;32m~\\AppData\\Local\\Temp\\ipykernel_29952\\633183133.py\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[1;32m----> 1\u001b[1;33m \u001b[0mfilter_score\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0msort_values\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mby\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;33m[\u001b[0m\u001b[1;34m'filter score'\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mascending\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0minplace\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m      2\u001b[0m \u001b[0mfilter_score\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mreset_index\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdrop\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0minplace\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mNameError\u001b[0m: name 'filter_score' is not defined"
     ]
    }
   ],
   "source": [
    "filter_score.sort_values(by=['filter score'], ascending=False, inplace=True)\n",
    "filter_score.reset_index(drop=True,inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filter_score.head(30)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filter_score.tail(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "filter_score.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "filter_score.head(80).to_csv('filter_top.csv')\n",
    "vars_keep = list(filter_score['variable'][num_files:num_filter+3])\n",
    "print(i,' vars_keep:',vars_keep)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "vars_keep_df = pd.DataFrame({'col':vars_keep})\n",
    "vars_keep_df.to_csv('vars_keep_filter.csv',index=False)\n",
    "df_keep = df_top.filter(vars_keep, axis=1)\n",
    "df_keep.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_keep.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y = Y.values\n",
    "Y_save = Y.copy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Y = np.array(Y)\n",
    "X = df_keep\n",
    "print(Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('time to here:', dt.datetime.now() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(X.shape,Y.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(type(X),type(Y))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# I'd like to define a scoring for the wrapper that's KS, but I haven't gotten around to this yet.\n",
    "# def KSscore(classifier, x,y)\n",
    "#     goods = "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def fdr(classifier, x, y, cutoff=detect_rate):\n",
    "# Calculates FDR score for the given classifier on dataset x and y with cutoff value\n",
    "# get the probability list from the given classifier\n",
    "    return fdr_prob(y, classifier.predict_proba(x), cutoff)\n",
    "def fdr_prob(y, y_prob, cutoff=detect_rate):\n",
    "    if len(y_prob.shape) != 1:    # sometimes the proba list can contain many columns, one for each category\n",
    "        y_prob = y_prob[:, -1:]   # only the last one (fraud_label==1) is used here.\n",
    "    num_fraud = len(y[y == 1])    # count the total nunber of frauds   \n",
    "# sort the proba list from high to low while retain the true (not predicted) fraud label\n",
    "    sorted_prob = np.asarray(sorted(zip(y_prob, y), key=lambda x: x[0], reverse=True))\n",
    "    cutoff_bin = sorted_prob[0:int(len(y) * cutoff), 1:]  # 3% cutoff\n",
    "# return the FDR score (#fraud_in_cutoff / #total_fraud)\n",
    "    return len(cutoff_bin[cutoff_bin == 1]) / num_fraud   "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Run a wrapper on the remaining top variables"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "%%time\n",
    "import warnings \n",
    "warnings.filterwarnings(\"ignore\", category=DeprecationWarning) \n",
    "# If you're doing forward selection it's enough to stop at num_wrapper variables. \n",
    "# If you're doing backward selection you need to go through all the variables to get a sorted list of num_wrapper variables.\n",
    "\n",
    "# I can't figure out how to get rid of this annoying warning! I don't know what I'm doing wrong...\n",
    "\n",
    "nfeatures = len(X.columns)\n",
    "# clf = RandomForestClassifier(n_estimators=5) # simple, fast nonlinear model for the wrapper\n",
    "clf = LGBMClassifier(n_estimators=40,num_leaves=4) # simple, fast nonlinear model for the wrapper\n",
    "sfs = SFS(clf,k_features=num_wrapper,forward=True,verbose=0,scoring=fdr,cv=4,n_jobs=-1) # use for forward selection\n",
    "# sfs = SFS(clf,k_features=1,forward=False,verbose=0,scoring=fdr,cv=4,n_jobs=-1) # use for backward selection\n",
    "sfs.fit(X,Y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "time to here: 0:10:29.394128\n"
     ]
    }
   ],
   "source": [
    "print('time to here:', dt.datetime.now() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1500x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "from mlxtend.plotting import plot_sequential_feature_selection as plot_sfs\n",
    "fig1 = plot_sfs(sfs.get_metric_dict(),kind='std_dev', figsize=(15, 6))\n",
    "# plt.xticks(np.arange(0, len(X.columns), step=5))\n",
    "plt.xticks(np.arange(0, num_wrapper, step=5))\n",
    "plt.yticks(np.arange(0,1,step=.1))\n",
    "plt.ylim([.5, .8])\n",
    "plt.xlim(0,num_wrapper)\n",
    "plt.title('Stepwise Selection')\n",
    "plt.grid()\n",
    "plt.savefig('performance_nvars.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "vars_FS = pd.DataFrame.from_dict(sfs.get_metric_dict()).T"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "ordered_vars_FS = vars_FS.copy()\n",
    "for i in range(len(ordered_vars_FS)):\n",
    "    ordered_vars_FS.loc[i+1,'add variables in this order'] = int(i+1)\n",
    "    if i+1 == 1:\n",
    "        ordered_vars_FS.loc[i+1,'variable name'] = (list(ordered_vars_FS.loc[i+1,'feature_names'])[0])\n",
    "    else:\n",
    "        ordered_vars_FS.loc[i+1,'variable name'] = (list(set(ordered_vars_FS.loc[i+1,'feature_names']) - set(ordered_vars_FS.loc[i,'feature_names'])))[0]\n",
    "# You might also need this following line. It converts a list to a string\n",
    "#         ordered_vars_FS.loc[i+1,'variable name'] = ordered_vars_FS.loc[i+1,'variable name'][0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>feature_idx</th>\n",
       "      <th>cv_scores</th>\n",
       "      <th>avg_score</th>\n",
       "      <th>feature_names</th>\n",
       "      <th>ci_bound</th>\n",
       "      <th>std_dev</th>\n",
       "      <th>std_err</th>\n",
       "      <th>add variables in this order</th>\n",
       "      <th>variable name</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>(15,)</td>\n",
       "      <td>[0.6908396946564885, 0.5670498084291188, 0.709...</td>\n",
       "      <td>0.595961</td>\n",
       "      <td>(card_merch_total_14,)</td>\n",
       "      <td>0.188308</td>\n",
       "      <td>0.117474</td>\n",
       "      <td>0.067823</td>\n",
       "      <td>1.0</td>\n",
       "      <td>card_merch_total_14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>(15, 16)</td>\n",
       "      <td>[0.7404580152671756, 0.6360153256704981, 0.744...</td>\n",
       "      <td>0.660912</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14)</td>\n",
       "      <td>0.145474</td>\n",
       "      <td>0.090752</td>\n",
       "      <td>0.052396</td>\n",
       "      <td>2.0</td>\n",
       "      <td>card_zip3_max_14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>(15, 16, 173)</td>\n",
       "      <td>[0.7595419847328244, 0.632183908045977, 0.7786...</td>\n",
       "      <td>0.676176</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_m...</td>\n",
       "      <td>0.159284</td>\n",
       "      <td>0.099367</td>\n",
       "      <td>0.05737</td>\n",
       "      <td>3.0</td>\n",
       "      <td>card_merch_avg_14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>(15, 16, 153, 173)</td>\n",
       "      <td>[0.7900763358778626, 0.6896551724137931, 0.797...</td>\n",
       "      <td>0.696269</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, state_...</td>\n",
       "      <td>0.187482</td>\n",
       "      <td>0.116958</td>\n",
       "      <td>0.067526</td>\n",
       "      <td>4.0</td>\n",
       "      <td>state_des_max_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>(15, 16, 45, 153, 173)</td>\n",
       "      <td>[0.767175572519084, 0.685823754789272, 0.82061...</td>\n",
       "      <td>0.701036</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.175552</td>\n",
       "      <td>0.109516</td>\n",
       "      <td>0.063229</td>\n",
       "      <td>5.0</td>\n",
       "      <td>card_zip_max_60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>(15, 16, 45, 153, 169, 173)</td>\n",
       "      <td>[0.7977099236641222, 0.6934865900383141, 0.812...</td>\n",
       "      <td>0.709631</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.17818</td>\n",
       "      <td>0.111155</td>\n",
       "      <td>0.064176</td>\n",
       "      <td>6.0</td>\n",
       "      <td>Card_Merchnum_desc_avg_30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>(15, 16, 45, 64, 153, 169, 173)</td>\n",
       "      <td>[0.7862595419847328, 0.6973180076628352, 0.824...</td>\n",
       "      <td>0.711543</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.176668</td>\n",
       "      <td>0.110212</td>\n",
       "      <td>0.063631</td>\n",
       "      <td>7.0</td>\n",
       "      <td>Card_Merchnum_Zip_total_60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>(15, 16, 45, 48, 64, 153, 169, 173)</td>\n",
       "      <td>[0.7748091603053435, 0.7049808429118773, 0.824...</td>\n",
       "      <td>0.709642</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.17591</td>\n",
       "      <td>0.10974</td>\n",
       "      <td>0.063358</td>\n",
       "      <td>8.0</td>\n",
       "      <td>Card_Merchdesc_Zip_total_30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>(15, 16, 45, 48, 64, 153, 169, 173, 192)</td>\n",
       "      <td>[0.8053435114503816, 0.6896551724137931, 0.816...</td>\n",
       "      <td>0.711536</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.182332</td>\n",
       "      <td>0.113746</td>\n",
       "      <td>0.065671</td>\n",
       "      <td>9.0</td>\n",
       "      <td>card_zip_avg_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>(15, 16, 45, 48, 64, 153, 169, 173, 174, 192)</td>\n",
       "      <td>[0.8053435114503816, 0.6819923371647509, 0.828...</td>\n",
       "      <td>0.713437</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.1851</td>\n",
       "      <td>0.115472</td>\n",
       "      <td>0.066668</td>\n",
       "      <td>10.0</td>\n",
       "      <td>Card_Merchnum_Zip_avg_14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>(15, 16, 45, 48, 64, 149, 153, 169, 173, 174, ...</td>\n",
       "      <td>[0.8053435114503816, 0.6934865900383141, 0.832...</td>\n",
       "      <td>0.716311</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.18789</td>\n",
       "      <td>0.117213</td>\n",
       "      <td>0.067673</td>\n",
       "      <td>11.0</td>\n",
       "      <td>card_merch_avg_3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 153, 169, 173, 1...</td>\n",
       "      <td>[0.8053435114503816, 0.6934865900383141, 0.839...</td>\n",
       "      <td>0.720127</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.186254</td>\n",
       "      <td>0.116192</td>\n",
       "      <td>0.067084</td>\n",
       "      <td>12.0</td>\n",
       "      <td>card_merch_total_60</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...</td>\n",
       "      <td>[0.8053435114503816, 0.6934865900383141, 0.839...</td>\n",
       "      <td>0.720127</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.186254</td>\n",
       "      <td>0.116192</td>\n",
       "      <td>0.067084</td>\n",
       "      <td>13.0</td>\n",
       "      <td>Card_Merchnum_Zip_avg_3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...</td>\n",
       "      <td>[0.7977099236641222, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.715356</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.179329</td>\n",
       "      <td>0.111872</td>\n",
       "      <td>0.06459</td>\n",
       "      <td>14.0</td>\n",
       "      <td>Card_Merchdesc_Zip_avg_7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...</td>\n",
       "      <td>[0.7977099236641222, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.715356</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.179329</td>\n",
       "      <td>0.111872</td>\n",
       "      <td>0.06459</td>\n",
       "      <td>15.0</td>\n",
       "      <td>Card_Merchdesc_avg_7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...</td>\n",
       "      <td>[0.8015267175572519, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.716311</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.180471</td>\n",
       "      <td>0.112585</td>\n",
       "      <td>0.065001</td>\n",
       "      <td>16.0</td>\n",
       "      <td>Card_Merchnum_Zip_avg_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 149, 151, 153, 168, 1...</td>\n",
       "      <td>[0.8015267175572519, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.716311</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.180471</td>\n",
       "      <td>0.112585</td>\n",
       "      <td>0.065001</td>\n",
       "      <td>17.0</td>\n",
       "      <td>Card_Merchnum_desc_avg_7</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 16...</td>\n",
       "      <td>[0.7977099236641222, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.717265</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.174605</td>\n",
       "      <td>0.108925</td>\n",
       "      <td>0.062888</td>\n",
       "      <td>18.0</td>\n",
       "      <td>Card_Merchnum_desc_max_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 16...</td>\n",
       "      <td>[0.7977099236641222, 0.6934865900383141, 0.828...</td>\n",
       "      <td>0.717265</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.174605</td>\n",
       "      <td>0.108925</td>\n",
       "      <td>0.062888</td>\n",
       "      <td>19.0</td>\n",
       "      <td>card_merch_avg_1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>(15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 15...</td>\n",
       "      <td>[0.7977099236641222, 0.7011494252873564, 0.824...</td>\n",
       "      <td>0.714409</td>\n",
       "      <td>(card_merch_total_14, card_zip3_max_14, card_z...</td>\n",
       "      <td>0.182134</td>\n",
       "      <td>0.113622</td>\n",
       "      <td>0.0656</td>\n",
       "      <td>20.0</td>\n",
       "      <td>merch_state_max_1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                          feature_idx  \\\n",
       "1                                               (15,)   \n",
       "2                                            (15, 16)   \n",
       "3                                       (15, 16, 173)   \n",
       "4                                  (15, 16, 153, 173)   \n",
       "5                              (15, 16, 45, 153, 173)   \n",
       "6                         (15, 16, 45, 153, 169, 173)   \n",
       "7                     (15, 16, 45, 64, 153, 169, 173)   \n",
       "8                 (15, 16, 45, 48, 64, 153, 169, 173)   \n",
       "9            (15, 16, 45, 48, 64, 153, 169, 173, 192)   \n",
       "10      (15, 16, 45, 48, 64, 153, 169, 173, 174, 192)   \n",
       "11  (15, 16, 45, 48, 64, 149, 153, 169, 173, 174, ...   \n",
       "12  (15, 16, 45, 48, 62, 64, 149, 153, 169, 173, 1...   \n",
       "13  (15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...   \n",
       "14  (15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...   \n",
       "15  (15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...   \n",
       "16  (15, 16, 45, 48, 62, 64, 149, 151, 153, 169, 1...   \n",
       "17  (15, 16, 45, 48, 62, 64, 149, 151, 153, 168, 1...   \n",
       "18  (15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 16...   \n",
       "19  (15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 16...   \n",
       "20  (15, 16, 45, 48, 62, 64, 96, 149, 151, 153, 15...   \n",
       "\n",
       "                                            cv_scores avg_score  \\\n",
       "1   [0.6908396946564885, 0.5670498084291188, 0.709...  0.595961   \n",
       "2   [0.7404580152671756, 0.6360153256704981, 0.744...  0.660912   \n",
       "3   [0.7595419847328244, 0.632183908045977, 0.7786...  0.676176   \n",
       "4   [0.7900763358778626, 0.6896551724137931, 0.797...  0.696269   \n",
       "5   [0.767175572519084, 0.685823754789272, 0.82061...  0.701036   \n",
       "6   [0.7977099236641222, 0.6934865900383141, 0.812...  0.709631   \n",
       "7   [0.7862595419847328, 0.6973180076628352, 0.824...  0.711543   \n",
       "8   [0.7748091603053435, 0.7049808429118773, 0.824...  0.709642   \n",
       "9   [0.8053435114503816, 0.6896551724137931, 0.816...  0.711536   \n",
       "10  [0.8053435114503816, 0.6819923371647509, 0.828...  0.713437   \n",
       "11  [0.8053435114503816, 0.6934865900383141, 0.832...  0.716311   \n",
       "12  [0.8053435114503816, 0.6934865900383141, 0.839...  0.720127   \n",
       "13  [0.8053435114503816, 0.6934865900383141, 0.839...  0.720127   \n",
       "14  [0.7977099236641222, 0.6934865900383141, 0.828...  0.715356   \n",
       "15  [0.7977099236641222, 0.6934865900383141, 0.828...  0.715356   \n",
       "16  [0.8015267175572519, 0.6934865900383141, 0.828...  0.716311   \n",
       "17  [0.8015267175572519, 0.6934865900383141, 0.828...  0.716311   \n",
       "18  [0.7977099236641222, 0.6934865900383141, 0.828...  0.717265   \n",
       "19  [0.7977099236641222, 0.6934865900383141, 0.828...  0.717265   \n",
       "20  [0.7977099236641222, 0.7011494252873564, 0.824...  0.714409   \n",
       "\n",
       "                                        feature_names  ci_bound   std_dev  \\\n",
       "1                              (card_merch_total_14,)  0.188308  0.117474   \n",
       "2             (card_merch_total_14, card_zip3_max_14)  0.145474  0.090752   \n",
       "3   (card_merch_total_14, card_zip3_max_14, card_m...  0.159284  0.099367   \n",
       "4   (card_merch_total_14, card_zip3_max_14, state_...  0.187482  0.116958   \n",
       "5   (card_merch_total_14, card_zip3_max_14, card_z...  0.175552  0.109516   \n",
       "6   (card_merch_total_14, card_zip3_max_14, card_z...   0.17818  0.111155   \n",
       "7   (card_merch_total_14, card_zip3_max_14, card_z...  0.176668  0.110212   \n",
       "8   (card_merch_total_14, card_zip3_max_14, card_z...   0.17591   0.10974   \n",
       "9   (card_merch_total_14, card_zip3_max_14, card_z...  0.182332  0.113746   \n",
       "10  (card_merch_total_14, card_zip3_max_14, card_z...    0.1851  0.115472   \n",
       "11  (card_merch_total_14, card_zip3_max_14, card_z...   0.18789  0.117213   \n",
       "12  (card_merch_total_14, card_zip3_max_14, card_z...  0.186254  0.116192   \n",
       "13  (card_merch_total_14, card_zip3_max_14, card_z...  0.186254  0.116192   \n",
       "14  (card_merch_total_14, card_zip3_max_14, card_z...  0.179329  0.111872   \n",
       "15  (card_merch_total_14, card_zip3_max_14, card_z...  0.179329  0.111872   \n",
       "16  (card_merch_total_14, card_zip3_max_14, card_z...  0.180471  0.112585   \n",
       "17  (card_merch_total_14, card_zip3_max_14, card_z...  0.180471  0.112585   \n",
       "18  (card_merch_total_14, card_zip3_max_14, card_z...  0.174605  0.108925   \n",
       "19  (card_merch_total_14, card_zip3_max_14, card_z...  0.174605  0.108925   \n",
       "20  (card_merch_total_14, card_zip3_max_14, card_z...  0.182134  0.113622   \n",
       "\n",
       "     std_err  add variables in this order                variable name  \n",
       "1   0.067823                          1.0          card_merch_total_14  \n",
       "2   0.052396                          2.0             card_zip3_max_14  \n",
       "3    0.05737                          3.0            card_merch_avg_14  \n",
       "4   0.067526                          4.0              state_des_max_1  \n",
       "5   0.063229                          5.0              card_zip_max_60  \n",
       "6   0.064176                          6.0    Card_Merchnum_desc_avg_30  \n",
       "7   0.063631                          7.0   Card_Merchnum_Zip_total_60  \n",
       "8   0.063358                          8.0  Card_Merchdesc_Zip_total_30  \n",
       "9   0.065671                          9.0               card_zip_avg_1  \n",
       "10  0.066668                         10.0     Card_Merchnum_Zip_avg_14  \n",
       "11  0.067673                         11.0             card_merch_avg_3  \n",
       "12  0.067084                         12.0          card_merch_total_60  \n",
       "13  0.067084                         13.0      Card_Merchnum_Zip_avg_3  \n",
       "14   0.06459                         14.0     Card_Merchdesc_Zip_avg_7  \n",
       "15   0.06459                         15.0         Card_Merchdesc_avg_7  \n",
       "16  0.065001                         16.0      Card_Merchnum_Zip_avg_1  \n",
       "17  0.065001                         17.0     Card_Merchnum_desc_avg_7  \n",
       "18  0.062888                         18.0     Card_Merchnum_desc_max_1  \n",
       "19  0.062888                         19.0             card_merch_avg_1  \n",
       "20    0.0656                         20.0            merch_state_max_1  "
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ordered_vars_FS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "ordered_vars_FS.to_csv('Wrapper_selection_info.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "1             card_merch_total_14\n",
       "2                card_zip3_max_14\n",
       "3               card_merch_avg_14\n",
       "4                 state_des_max_1\n",
       "5                 card_zip_max_60\n",
       "6       Card_Merchnum_desc_avg_30\n",
       "7      Card_Merchnum_Zip_total_60\n",
       "8     Card_Merchdesc_Zip_total_30\n",
       "9                  card_zip_avg_1\n",
       "10       Card_Merchnum_Zip_avg_14\n",
       "11               card_merch_avg_3\n",
       "12            card_merch_total_60\n",
       "13        Card_Merchnum_Zip_avg_3\n",
       "14       Card_Merchdesc_Zip_avg_7\n",
       "15           Card_Merchdesc_avg_7\n",
       "16        Card_Merchnum_Zip_avg_1\n",
       "17       Card_Merchnum_desc_avg_7\n",
       "18       Card_Merchnum_desc_max_1\n",
       "19               card_merch_avg_1\n",
       "20              merch_state_max_1\n",
       "Name: variable name, dtype: object"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vars_keep = ordered_vars_FS['variable name']\n",
    "vars_keep_list = ordered_vars_FS['variable name'].tolist()\n",
    "vars_keep.to_csv('final_vars_list.csv',index=False)\n",
    "vars_keep"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>filter score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>variable</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_7</th>\n",
       "      <td>0.676549</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip_total_7</th>\n",
       "      <td>0.666816</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_3</th>\n",
       "      <td>0.660260</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_14</th>\n",
       "      <td>0.659257</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_state_total_7</th>\n",
       "      <td>0.654135</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>merch_zip_unique_count_for_merch_state_7</th>\n",
       "      <td>0.000022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Card_Merchdesc_Zip_unique_count_for_card_state_1</th>\n",
       "      <td>0.000022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_unique_count_for_card_state_1</th>\n",
       "      <td>0.000017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_merch_unique_count_for_card_state_1</th>\n",
       "      <td>0.000011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_merch_unique_count_for_Cardnum_1</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2343 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  filter score\n",
       "variable                                                      \n",
       "card_zip3_total_7                                     0.676549\n",
       "card_zip_total_7                                      0.666816\n",
       "card_zip3_total_3                                     0.660260\n",
       "card_zip3_total_14                                    0.659257\n",
       "card_state_total_7                                    0.654135\n",
       "...                                                        ...\n",
       "merch_zip_unique_count_for_merch_state_7              0.000022\n",
       "Card_Merchdesc_Zip_unique_count_for_card_state_1      0.000022\n",
       "card_zip3_unique_count_for_card_state_1               0.000017\n",
       "card_merch_unique_count_for_card_state_1              0.000011\n",
       "card_merch_unique_count_for_Cardnum_1                 0.000000\n",
       "\n",
       "[2343 rows x 1 columns]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "filter_score.set_index('variable',drop=True,inplace=True)\n",
    "filter_score = filter_score.iloc[1:,:]\n",
    "filter_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>variable</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>card_merch_total_14</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_max_14</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_merch_avg_14</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>state_des_max_1</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip_max_60</th>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "Empty DataFrame\n",
       "Columns: []\n",
       "Index: [card_merch_total_14, card_zip3_max_14, card_merch_avg_14, state_des_max_1, card_zip_max_60]"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vars_keep_sorted = pd.DataFrame(vars_keep_list)\n",
    "vars_keep_sorted.columns=['variable']\n",
    "vars_keep_sorted.set_index('variable',drop=True,inplace=True)\n",
    "vars_keep_sorted.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "vars_keep_sorted = pd.concat([vars_keep_sorted,filter_score],axis=1,join='inner')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>wrapper order</th>\n",
       "      <th>variable</th>\n",
       "      <th>filter score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>card_merch_total_14</td>\n",
       "      <td>0.630048</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>card_zip3_max_14</td>\n",
       "      <td>0.629515</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>3</td>\n",
       "      <td>card_merch_avg_14</td>\n",
       "      <td>0.518386</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>4</td>\n",
       "      <td>state_des_max_1</td>\n",
       "      <td>0.524301</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>5</td>\n",
       "      <td>card_zip_max_60</td>\n",
       "      <td>0.605193</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>6</td>\n",
       "      <td>Card_Merchnum_desc_avg_30</td>\n",
       "      <td>0.519394</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>7</td>\n",
       "      <td>Card_Merchnum_Zip_total_60</td>\n",
       "      <td>0.597034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>8</td>\n",
       "      <td>Card_Merchdesc_Zip_total_30</td>\n",
       "      <td>0.604455</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>9</td>\n",
       "      <td>card_zip_avg_1</td>\n",
       "      <td>0.514891</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>10</td>\n",
       "      <td>Card_Merchnum_Zip_avg_14</td>\n",
       "      <td>0.518122</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>11</td>\n",
       "      <td>card_merch_avg_3</td>\n",
       "      <td>0.525502</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>12</td>\n",
       "      <td>card_merch_total_60</td>\n",
       "      <td>0.598597</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>13</td>\n",
       "      <td>Card_Merchnum_Zip_avg_3</td>\n",
       "      <td>0.524558</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>14</td>\n",
       "      <td>Card_Merchdesc_Zip_avg_7</td>\n",
       "      <td>0.517607</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>15</td>\n",
       "      <td>Card_Merchdesc_avg_7</td>\n",
       "      <td>0.516608</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>16</td>\n",
       "      <td>Card_Merchnum_Zip_avg_1</td>\n",
       "      <td>0.514052</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>17</td>\n",
       "      <td>Card_Merchnum_desc_avg_7</td>\n",
       "      <td>0.519505</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>18</td>\n",
       "      <td>Card_Merchnum_desc_max_1</td>\n",
       "      <td>0.560029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>19</td>\n",
       "      <td>card_merch_avg_1</td>\n",
       "      <td>0.515008</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>20</td>\n",
       "      <td>merch_state_max_1</td>\n",
       "      <td>0.522978</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    wrapper order                     variable  filter score\n",
       "0               1          card_merch_total_14      0.630048\n",
       "1               2             card_zip3_max_14      0.629515\n",
       "2               3            card_merch_avg_14      0.518386\n",
       "3               4              state_des_max_1      0.524301\n",
       "4               5              card_zip_max_60      0.605193\n",
       "5               6    Card_Merchnum_desc_avg_30      0.519394\n",
       "6               7   Card_Merchnum_Zip_total_60      0.597034\n",
       "7               8  Card_Merchdesc_Zip_total_30      0.604455\n",
       "8               9               card_zip_avg_1      0.514891\n",
       "9              10     Card_Merchnum_Zip_avg_14      0.518122\n",
       "10             11             card_merch_avg_3      0.525502\n",
       "11             12          card_merch_total_60      0.598597\n",
       "12             13      Card_Merchnum_Zip_avg_3      0.524558\n",
       "13             14     Card_Merchdesc_Zip_avg_7      0.517607\n",
       "14             15         Card_Merchdesc_avg_7      0.516608\n",
       "15             16      Card_Merchnum_Zip_avg_1      0.514052\n",
       "16             17     Card_Merchnum_desc_avg_7      0.519505\n",
       "17             18     Card_Merchnum_desc_max_1      0.560029\n",
       "18             19             card_merch_avg_1      0.515008\n",
       "19             20            merch_state_max_1      0.522978"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vars_keep_sorted.reset_index(inplace=True)\n",
    "vars_keep_sorted.reset_index(inplace=True)\n",
    "vars_keep_sorted['index'] = vars_keep_sorted['index'] + 1\n",
    "vars_keep_sorted.rename(columns={'index':'wrapper order'},inplace=True)\n",
    "vars_keep_sorted.to_csv('vars_keep_sorted.csv',index=False)\n",
    "vars_keep_sorted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['card_merch_total_14',\n",
       " 'card_zip3_max_14',\n",
       " 'card_merch_avg_14',\n",
       " 'state_des_max_1',\n",
       " 'card_zip_max_60',\n",
       " 'Card_Merchnum_desc_avg_30',\n",
       " 'Card_Merchnum_Zip_total_60',\n",
       " 'Card_Merchdesc_Zip_total_30',\n",
       " 'card_zip_avg_1',\n",
       " 'Card_Merchnum_Zip_avg_14',\n",
       " 'card_merch_avg_3',\n",
       " 'card_merch_total_60',\n",
       " 'Card_Merchnum_Zip_avg_3',\n",
       " 'Card_Merchdesc_Zip_avg_7',\n",
       " 'Card_Merchdesc_avg_7',\n",
       " 'Card_Merchnum_Zip_avg_1',\n",
       " 'Card_Merchnum_desc_avg_7',\n",
       " 'Card_Merchnum_desc_max_1',\n",
       " 'card_merch_avg_1',\n",
       " 'merch_state_max_1',\n",
       " 'Recnum',\n",
       " 'Fraud']"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vars_keep_list.append(index_name)\n",
    "vars_keep_list.append(y_name)\n",
    "vars_keep_list"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>filter score</th>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>variable</th>\n",
       "      <th></th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_7</th>\n",
       "      <td>0.676549</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip_total_7</th>\n",
       "      <td>0.666816</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_3</th>\n",
       "      <td>0.660260</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_total_14</th>\n",
       "      <td>0.659257</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_state_total_7</th>\n",
       "      <td>0.654135</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>merch_zip_unique_count_for_merch_state_7</th>\n",
       "      <td>0.000022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>Card_Merchdesc_Zip_unique_count_for_card_state_1</th>\n",
       "      <td>0.000022</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_zip3_unique_count_for_card_state_1</th>\n",
       "      <td>0.000017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_merch_unique_count_for_card_state_1</th>\n",
       "      <td>0.000011</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>card_merch_unique_count_for_Cardnum_1</th>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>2343 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  filter score\n",
       "variable                                                      \n",
       "card_zip3_total_7                                     0.676549\n",
       "card_zip_total_7                                      0.666816\n",
       "card_zip3_total_3                                     0.660260\n",
       "card_zip3_total_14                                    0.659257\n",
       "card_state_total_7                                    0.654135\n",
       "...                                                        ...\n",
       "merch_zip_unique_count_for_merch_state_7              0.000022\n",
       "Card_Merchdesc_Zip_unique_count_for_card_state_1      0.000022\n",
       "card_zip3_unique_count_for_card_state_1               0.000017\n",
       "card_merch_unique_count_for_card_state_1              0.000011\n",
       "card_merch_unique_count_for_Cardnum_1                 0.000000\n",
       "\n",
       "[2343 rows x 1 columns]"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "filter_score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: user 12.5 s, sys: 1.78 s, total: 14.3 s\n",
      "Wall time: 14.5 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(96397, 2344)"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "df = pd.read_csv(file_name)\n",
    "df.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(96397, 22)\n"
     ]
    }
   ],
   "source": [
    "df_keep = df.filter(vars_keep_list, axis=1)\n",
    "# df_keep = df[df.index.isin(vars_keep_list)]\n",
    "print(df_keep.shape)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_keep.to_csv('vars_final.csv',index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "duration:  0:10:44.857541\n"
     ]
    }
   ],
   "source": [
    "print(\"duration: \", dt.datetime.now() - start_time)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/Users/stevecoggeshall/Documents/Teaching/Data sets/done/transactions/transactions 2023 new'"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "anaconda-cloud": {},
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}