Preprocessing_model_tuning.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6097e757",
   "metadata": {},
   "source": [
    "# AKI - early/delayed renal treatment"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "db3cec7f",
   "metadata": {},
   "source": [
    "*import libraries*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "890c5c94",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Import libraries here\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib\n",
    "import seaborn as sns\n",
    "pd.options.mode.chained_assignment = None  # default='warn'\n",
    "\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.preprocessing import StandardScaler, PolynomialFeatures\n",
    "from sklearn.linear_model import (RidgeCV, \n",
    "                                  LassoCV, \n",
    "                                  ElasticNetCV, \n",
    "                                  LogisticRegressionCV,\n",
    "                                  LinearRegression,\n",
    "                                  LogisticRegression)\n",
    "from imblearn.over_sampling import SMOTE\n",
    "from imblearn.pipeline import Pipeline\n",
    "from sklearn.metrics import (confusion_matrix, \n",
    "                             plot_confusion_matrix, \n",
    "                             roc_auc_score, \n",
    "                             plot_roc_curve, \n",
    "                             accuracy_score,\n",
    "                            )\n",
    "from sklearn.decomposition import PCA\n",
    "\n",
    "from sklearn.linear_model import LogisticRegression\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier, ExtraTreesClassifier, BaggingClassifier, AdaBoostClassifier\n",
    "from sklearn.svm import LinearSVC, SVC\n",
    "\n",
    "import xgboost as xgb\n",
    "import time\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "5beeeb19",
   "metadata": {},
   "outputs": [],
   "source": [
    "# import data\n",
    "aki_data = pd.read_csv('cleaned_data.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "8be77db3",
   "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>weight</th>\n",
       "      <th>aki_stage</th>\n",
       "      <th>delay_rrt</th>\n",
       "      <th>gender</th>\n",
       "      <th>admission_age</th>\n",
       "      <th>ethnicity</th>\n",
       "      <th>hematocrit_min</th>\n",
       "      <th>hematocrit_max</th>\n",
       "      <th>hemoglobin_min</th>\n",
       "      <th>hemoglobin_max</th>\n",
       "      <th>platelets_min</th>\n",
       "      <th>platelets_max</th>\n",
       "      <th>wbc_min</th>\n",
       "      <th>wbc_max</th>\n",
       "      <th>aniongap_min</th>\n",
       "      <th>aniongap_max</th>\n",
       "      <th>bicarbonate_min</th>\n",
       "      <th>bicarbonate_max</th>\n",
       "      <th>bun_min</th>\n",
       "      <th>bun_max</th>\n",
       "      <th>calcium_min</th>\n",
       "      <th>calcium_max</th>\n",
       "      <th>chloride_min</th>\n",
       "      <th>chloride_max</th>\n",
       "      <th>creatinine_min</th>\n",
       "      <th>creatinine_max</th>\n",
       "      <th>glucose_min</th>\n",
       "      <th>glucose_max</th>\n",
       "      <th>sodium_min</th>\n",
       "      <th>sodium_max</th>\n",
       "      <th>potassium_min</th>\n",
       "      <th>potassium_max</th>\n",
       "      <th>inr_max</th>\n",
       "      <th>pt_max</th>\n",
       "      <th>ptt_max</th>\n",
       "      <th>heart_rate_mean</th>\n",
       "      <th>sbp_mean</th>\n",
       "      <th>dbp_mean</th>\n",
       "      <th>mbp_mean</th>\n",
       "      <th>resp_rate_mean</th>\n",
       "      <th>temperature_mean</th>\n",
       "      <th>spo2_mean</th>\n",
       "      <th>glucose_mean</th>\n",
       "      <th>gcs_min</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>104.50</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>66.262081</td>\n",
       "      <td>unknown</td>\n",
       "      <td>31.4</td>\n",
       "      <td>31.9</td>\n",
       "      <td>11.3</td>\n",
       "      <td>11.3</td>\n",
       "      <td>246.0</td>\n",
       "      <td>247.0</td>\n",
       "      <td>11.2</td>\n",
       "      <td>11.2</td>\n",
       "      <td>20.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>24.0</td>\n",
       "      <td>109.0</td>\n",
       "      <td>114.0</td>\n",
       "      <td>8.7</td>\n",
       "      <td>9.5</td>\n",
       "      <td>106.0</td>\n",
       "      <td>114.0</td>\n",
       "      <td>3.0</td>\n",
       "      <td>3.1</td>\n",
       "      <td>172.0</td>\n",
       "      <td>317.0</td>\n",
       "      <td>146.0</td>\n",
       "      <td>154.0</td>\n",
       "      <td>3.9</td>\n",
       "      <td>4.1</td>\n",
       "      <td>1.3</td>\n",
       "      <td>14.6</td>\n",
       "      <td>47.5</td>\n",
       "      <td>89.625000</td>\n",
       "      <td>132.500000</td>\n",
       "      <td>68.5000</td>\n",
       "      <td>83.090909</td>\n",
       "      <td>21.673077</td>\n",
       "      <td>37.405000</td>\n",
       "      <td>96.909091</td>\n",
       "      <td>257.500000</td>\n",
       "      <td>10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>100.00</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>70.489938</td>\n",
       "      <td>black</td>\n",
       "      <td>34.0</td>\n",
       "      <td>34.2</td>\n",
       "      <td>11.3</td>\n",
       "      <td>11.4</td>\n",
       "      <td>185.0</td>\n",
       "      <td>186.0</td>\n",
       "      <td>12.1</td>\n",
       "      <td>13.5</td>\n",
       "      <td>14.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>20.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>36.0</td>\n",
       "      <td>41.0</td>\n",
       "      <td>7.5</td>\n",
       "      <td>7.7</td>\n",
       "      <td>103.0</td>\n",
       "      <td>104.0</td>\n",
       "      <td>1.1</td>\n",
       "      <td>1.1</td>\n",
       "      <td>106.0</td>\n",
       "      <td>140.0</td>\n",
       "      <td>134.0</td>\n",
       "      <td>135.0</td>\n",
       "      <td>4.4</td>\n",
       "      <td>4.8</td>\n",
       "      <td>1.3</td>\n",
       "      <td>13.9</td>\n",
       "      <td>25.0</td>\n",
       "      <td>84.208333</td>\n",
       "      <td>107.500000</td>\n",
       "      <td>56.4600</td>\n",
       "      <td>74.019231</td>\n",
       "      <td>17.134615</td>\n",
       "      <td>36.978000</td>\n",
       "      <td>96.680000</td>\n",
       "      <td>123.666667</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>98.95</td>\n",
       "      <td>3</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>37.700917</td>\n",
       "      <td>unknown</td>\n",
       "      <td>27.1</td>\n",
       "      <td>29.4</td>\n",
       "      <td>9.9</td>\n",
       "      <td>10.4</td>\n",
       "      <td>185.0</td>\n",
       "      <td>252.0</td>\n",
       "      <td>22.4</td>\n",
       "      <td>31.1</td>\n",
       "      <td>26.0</td>\n",
       "      <td>30.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>40.0</td>\n",
       "      <td>53.0</td>\n",
       "      <td>7.4</td>\n",
       "      <td>10.2</td>\n",
       "      <td>73.0</td>\n",
       "      <td>76.0</td>\n",
       "      <td>7.8</td>\n",
       "      <td>10.5</td>\n",
       "      <td>121.0</td>\n",
       "      <td>228.0</td>\n",
       "      <td>118.0</td>\n",
       "      <td>119.0</td>\n",
       "      <td>3.3</td>\n",
       "      <td>3.8</td>\n",
       "      <td>2.6</td>\n",
       "      <td>27.5</td>\n",
       "      <td>36.0</td>\n",
       "      <td>80.956522</td>\n",
       "      <td>101.958333</td>\n",
       "      <td>52.4375</td>\n",
       "      <td>67.541667</td>\n",
       "      <td>20.000000</td>\n",
       "      <td>36.132143</td>\n",
       "      <td>95.250000</td>\n",
       "      <td>152.625000</td>\n",
       "      <td>6</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>105.70</td>\n",
       "      <td>3</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>59.967156</td>\n",
       "      <td>white</td>\n",
       "      <td>25.0</td>\n",
       "      <td>30.9</td>\n",
       "      <td>8.1</td>\n",
       "      <td>10.1</td>\n",
       "      <td>70.0</td>\n",
       "      <td>107.0</td>\n",
       "      <td>7.6</td>\n",
       "      <td>14.5</td>\n",
       "      <td>11.0</td>\n",
       "      <td>13.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>23.0</td>\n",
       "      <td>15.0</td>\n",
       "      <td>16.0</td>\n",
       "      <td>7.5</td>\n",
       "      <td>7.8</td>\n",
       "      <td>106.0</td>\n",
       "      <td>110.0</td>\n",
       "      <td>1.1</td>\n",
       "      <td>1.2</td>\n",
       "      <td>47.0</td>\n",
       "      <td>127.0</td>\n",
       "      <td>140.0</td>\n",
       "      <td>141.0</td>\n",
       "      <td>3.2</td>\n",
       "      <td>4.0</td>\n",
       "      <td>2.6</td>\n",
       "      <td>28.4</td>\n",
       "      <td>51.3</td>\n",
       "      <td>87.521739</td>\n",
       "      <td>108.000000</td>\n",
       "      <td>56.4000</td>\n",
       "      <td>70.700000</td>\n",
       "      <td>24.086957</td>\n",
       "      <td>36.776667</td>\n",
       "      <td>96.478261</td>\n",
       "      <td>102.714286</td>\n",
       "      <td>13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>69.00</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>79.060470</td>\n",
       "      <td>white</td>\n",
       "      <td>38.0</td>\n",
       "      <td>38.0</td>\n",
       "      <td>13.2</td>\n",
       "      <td>13.2</td>\n",
       "      <td>262.0</td>\n",
       "      <td>262.0</td>\n",
       "      <td>10.8</td>\n",
       "      <td>10.8</td>\n",
       "      <td>18.0</td>\n",
       "      <td>18.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>21.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>19.0</td>\n",
       "      <td>9.4</td>\n",
       "      <td>9.4</td>\n",
       "      <td>106.0</td>\n",
       "      <td>106.0</td>\n",
       "      <td>1.3</td>\n",
       "      <td>1.3</td>\n",
       "      <td>126.0</td>\n",
       "      <td>126.0</td>\n",
       "      <td>141.0</td>\n",
       "      <td>141.0</td>\n",
       "      <td>3.7</td>\n",
       "      <td>3.7</td>\n",
       "      <td>1.2</td>\n",
       "      <td>13.8</td>\n",
       "      <td>35.6</td>\n",
       "      <td>90.318182</td>\n",
       "      <td>140.800000</td>\n",
       "      <td>71.8500</td>\n",
       "      <td>88.285714</td>\n",
       "      <td>21.000000</td>\n",
       "      <td>36.620000</td>\n",
       "      <td>95.782609</td>\n",
       "      <td>124.000000</td>\n",
       "      <td>15</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   weight  aki_stage  delay_rrt  gender  admission_age ethnicity  \\\n",
       "0  104.50          1          1       1      66.262081   unknown   \n",
       "1  100.00          1          1       1      70.489938     black   \n",
       "2   98.95          3          0       1      37.700917   unknown   \n",
       "3  105.70          3          1       0      59.967156     white   \n",
       "4   69.00          1          0       1      79.060470     white   \n",
       "\n",
       "   hematocrit_min  hematocrit_max  hemoglobin_min  hemoglobin_max  \\\n",
       "0            31.4            31.9            11.3            11.3   \n",
       "1            34.0            34.2            11.3            11.4   \n",
       "2            27.1            29.4             9.9            10.4   \n",
       "3            25.0            30.9             8.1            10.1   \n",
       "4            38.0            38.0            13.2            13.2   \n",
       "\n",
       "   platelets_min  platelets_max  wbc_min  wbc_max  aniongap_min  aniongap_max  \\\n",
       "0          246.0          247.0     11.2     11.2          20.0          23.0   \n",
       "1          185.0          186.0     12.1     13.5          14.0          16.0   \n",
       "2          185.0          252.0     22.4     31.1          26.0          30.0   \n",
       "3           70.0          107.0      7.6     14.5          11.0          13.0   \n",
       "4          262.0          262.0     10.8     10.8          18.0          18.0   \n",
       "\n",
       "   bicarbonate_min  bicarbonate_max  bun_min  bun_max  calcium_min  \\\n",
       "0             23.0             24.0    109.0    114.0          8.7   \n",
       "1             20.0             21.0     36.0     41.0          7.5   \n",
       "2             15.0             18.0     40.0     53.0          7.4   \n",
       "3             19.0             23.0     15.0     16.0          7.5   \n",
       "4             21.0             21.0     19.0     19.0          9.4   \n",
       "\n",
       "   calcium_max  chloride_min  chloride_max  creatinine_min  creatinine_max  \\\n",
       "0          9.5         106.0         114.0             3.0             3.1   \n",
       "1          7.7         103.0         104.0             1.1             1.1   \n",
       "2         10.2          73.0          76.0             7.8            10.5   \n",
       "3          7.8         106.0         110.0             1.1             1.2   \n",
       "4          9.4         106.0         106.0             1.3             1.3   \n",
       "\n",
       "   glucose_min  glucose_max  sodium_min  sodium_max  potassium_min  \\\n",
       "0        172.0        317.0       146.0       154.0            3.9   \n",
       "1        106.0        140.0       134.0       135.0            4.4   \n",
       "2        121.0        228.0       118.0       119.0            3.3   \n",
       "3         47.0        127.0       140.0       141.0            3.2   \n",
       "4        126.0        126.0       141.0       141.0            3.7   \n",
       "\n",
       "   potassium_max  inr_max  pt_max  ptt_max  heart_rate_mean    sbp_mean  \\\n",
       "0            4.1      1.3    14.6     47.5        89.625000  132.500000   \n",
       "1            4.8      1.3    13.9     25.0        84.208333  107.500000   \n",
       "2            3.8      2.6    27.5     36.0        80.956522  101.958333   \n",
       "3            4.0      2.6    28.4     51.3        87.521739  108.000000   \n",
       "4            3.7      1.2    13.8     35.6        90.318182  140.800000   \n",
       "\n",
       "   dbp_mean   mbp_mean  resp_rate_mean  temperature_mean  spo2_mean  \\\n",
       "0   68.5000  83.090909       21.673077         37.405000  96.909091   \n",
       "1   56.4600  74.019231       17.134615         36.978000  96.680000   \n",
       "2   52.4375  67.541667       20.000000         36.132143  95.250000   \n",
       "3   56.4000  70.700000       24.086957         36.776667  96.478261   \n",
       "4   71.8500  88.285714       21.000000         36.620000  95.782609   \n",
       "\n",
       "   glucose_mean  gcs_min  \n",
       "0    257.500000       10  \n",
       "1    123.666667        3  \n",
       "2    152.625000        6  \n",
       "3    102.714286       13  \n",
       "4    124.000000       15  "
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pd.set_option('display.max_columns', 100)\n",
    "aki_data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "0dae0988",
   "metadata": {},
   "outputs": [],
   "source": [
    "# one-hot-encode categorical data ethnicity ?\n",
    "# drop ethnicity\n",
    "aki_data.drop(columns=['ethnicity'],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "96ceb15d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1778, 42)"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aki_data.drop(columns=['delay_rrt']).shape"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "253ad920",
   "metadata": {},
   "source": [
    "### Polynomial features "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d03e0aa5",
   "metadata": {},
   "source": [
    "Any of the features could be related to each other, or have an interactive effect with each other. Any synergistic effect on delay_rrt could increase the significant predictive power to the model. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "cc0eaa9a",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = aki_data.drop(columns=['delay_rrt'])\n",
    "y = aki_data['delay_rrt']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "b3d150b4",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generating the polynomial features table.  \n",
    "# instantiate\n",
    "poly = PolynomialFeatures(include_bias=False, degree=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "2ced0ef2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1778, 945)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# fit and transform the variables in the numerical dataframe\n",
    "X_poly = poly.fit_transform(X)\n",
    "X_poly.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "fee974ed",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Checking column names to all polynomial features\n",
    "X_poly = pd.DataFrame(X_poly,columns=poly.get_feature_names(X.columns))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "a2294fc9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Gennerating list of poly feature correlations\n",
    "X_poly_corrs = X_poly.corrwith(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "3f6b1811",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "chloride_min temperature_mean       0.198818\n",
       "chloride_min spo2_mean              0.182815\n",
       "chloride_min^2                      0.180279\n",
       "chloride_min                        0.179294\n",
       "bicarbonate_min chloride_min        0.158487\n",
       "bicarbonate_min chloride_max        0.152728\n",
       "chloride_min sodium_min             0.148421\n",
       "calcium_min chloride_min            0.145373\n",
       "chloride_min chloride_max           0.144399\n",
       "admission_age bicarbonate_min       0.140705\n",
       "admission_age chloride_min          0.137339\n",
       "bicarbonate_min temperature_mean    0.132319\n",
       "bicarbonate_min sodium_min          0.130960\n",
       "bicarbonate_min spo2_mean           0.130260\n",
       "bicarbonate_min                     0.128809\n",
       "bicarbonate_min sodium_max          0.126990\n",
       "bicarbonate_min heart_rate_mean     0.123727\n",
       "calcium_min chloride_max            0.121496\n",
       "chloride_max temperature_mean       0.121356\n",
       "chloride_min sodium_max             0.119685\n",
       "dtype: float64"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shows top 20 features most positively correlated with delay_rrt\n",
    "X_poly_corrs.sort_values(ascending=False).head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "4fbfe1ab",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "aniongap_max gcs_min            -0.248238\n",
       "aniongap_max temperature_mean   -0.249200\n",
       "aniongap_max sodium_max         -0.249964\n",
       "aniongap_max spo2_mean          -0.250221\n",
       "aniongap_max calcium_max        -0.254303\n",
       "aniongap_max                    -0.254689\n",
       "aniongap_max^2                  -0.255885\n",
       "creatinine_max resp_rate_mean   -0.256791\n",
       "creatinine_max gcs_min          -0.257502\n",
       "hemoglobin_max creatinine_max   -0.257905\n",
       "aniongap_min creatinine_max     -0.258068\n",
       "aniongap_max sbp_mean           -0.259088\n",
       "hematocrit_max creatinine_max   -0.259116\n",
       "calcium_max creatinine_max      -0.259650\n",
       "aniongap_min potassium_max      -0.260355\n",
       "aniongap_max dbp_mean           -0.262543\n",
       "creatinine_max potassium_max    -0.273823\n",
       "aniongap_max mbp_mean           -0.274197\n",
       "aniongap_max creatinine_max     -0.290806\n",
       "aniongap_max potassium_max      -0.291583\n",
       "dtype: float64"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Shows bottom 20 features most positively correlated with delay_rrt\n",
    "X_poly_corrs.sort_values(ascending=False).tail(20) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "114d573c",
   "metadata": {},
   "outputs": [],
   "source": [
    "# # Adding interaction features into train set ?\n",
    "# aki_data['aniongap_max*potassium_max'] = aki_data['aniongap_max'] * aki_data['potassium_max']\n",
    "# aki_data['calcium_max*creatinine_max'] = aki_data['calcium_max'] * aki_data['creatinine_max']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "75605b4a",
   "metadata": {},
   "outputs": [],
   "source": [
    "# X = aki_data.drop(columns=['delay_rrt'])\n",
    "# y = aki_data[['delay_rrt']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "463b3015",
   "metadata": {},
   "outputs": [],
   "source": [
    "# train test split\n",
    "# X_train, y_train : 50%\n",
    "# X_valid, y_valid : 25%\n",
    "# X_test, y_test   : 25%"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7c165d8b",
   "metadata": {},
   "source": [
    "### Baseline model (with SMOTE)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3969d4d9",
   "metadata": {},
   "source": [
    "From the correlated features, it seemed that 'aniongap_max' has the highest (negative) correlation. So will start with this. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "4297fc45",
   "metadata": {},
   "outputs": [],
   "source": [
    "X_sm = aki_data[['aniongap_max']]\n",
    "y_sm = aki_data[['delay_rrt']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "c66a38be",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "delay_rrt\n",
       "1            0.814961\n",
       "0            0.185039\n",
       "dtype: float64"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# imbalanced \n",
    "y_sm.value_counts(normalize=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "0c842565",
   "metadata": {},
   "outputs": [],
   "source": [
    "# split data\n",
    "Xsm_train, Xsm_test, ysm_train, ysm_test = train_test_split(X_sm, y_sm, test_size = 0.33, random_state=42, stratify =y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "37ed714e",
   "metadata": {},
   "outputs": [],
   "source": [
    "# scaling\n",
    "ss = StandardScaler()\n",
    "\n",
    "Xsm_train_sc = ss.fit_transform(Xsm_train)\n",
    "Xsm_test_sc = ss.transform(Xsm_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "b0efabfc",
   "metadata": {},
   "outputs": [],
   "source": [
    "# create synthetic data for training set\n",
    "smote = SMOTE()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "b0b01b25",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Xsmote_train, ysmote_train = smote.fit_sample(Xsm_train_sc, ysm_train)\n",
    "Xsmote_train, ysmote_train = smote.fit_resample(Xsm_train_sc, ysm_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "d885259d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<bound method SamplerMixin.fit of SMOTE()>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "smote.fit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "ba59a08b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "delay_rrt\n",
       "0            971\n",
       "1            971\n",
       "dtype: int64"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# balanced\n",
    "ysmote_train.value_counts()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "d6e3dfd2",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.6204943357363543, 0.6695059625212947)"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# instantiate\n",
    "lr = LogisticRegression()\n",
    "\n",
    "# fit model\n",
    "lr.fit(Xsmote_train, ysmote_train)\n",
    "\n",
    "# score\n",
    "lr.score(Xsmote_train, ysmote_train), lr.score(Xsm_test_sc, ysm_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "33975fd5",
   "metadata": {},
   "source": [
    "So the **baseline accuracy** is 0.620 (for train set), and 0.670 (for test set)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "76587191",
   "metadata": {},
   "source": [
    "## Feature selection"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "11f645d8",
   "metadata": {},
   "source": [
    "### Logistic regression (with SMOTE)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "736681ba",
   "metadata": {},
   "outputs": [],
   "source": [
    "# from previous notebook\n",
    "# drop the above columns\n",
    "aki_data.drop(columns=['gender','sbp_mean','dbp_mean'],inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "fc2ec91c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1778, 39)"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "aki_data.drop(columns=['delay_rrt']).shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "64876480",
   "metadata": {},
   "outputs": [],
   "source": [
    "# using (most) entire dataset WITHOUT polynomic features\n",
    "X = aki_data.drop(columns=['delay_rrt'])\n",
    "y = aki_data[['delay_rrt']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "061bd496",
   "metadata": {},
   "outputs": [],
   "source": [
    "# split data\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.33, random_state = 42, stratify =y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "id": "2922c73e",
   "metadata": {},
   "outputs": [],
   "source": [
    "pipe = Pipeline([\n",
    "        ('scale', StandardScaler()),\n",
    "        ('sampling', SMOTE(sampling_strategy='minority')),\n",
    "        ('logreg', LogisticRegression(max_iter=1_000, solver='saga'))\n",
    "    ])\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "3f23e79f",
   "metadata": {},
   "outputs": [],
   "source": [
    "# regardless of parameters, there is overfitting.\n",
    "pipe_params = {\n",
    "    'sampling__k_neighbors': [5, 10],   # \n",
    "    'logreg__penalty': ['l2', 'l1'],   # \n",
    "    'logreg__C': [1] # \n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "5b9d9e1d",
   "metadata": {},
   "outputs": [],
   "source": [
    "grid = GridSearchCV(pipe, pipe_params, scoring='roc_auc', n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "a2aea3a5",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(estimator=Pipeline(steps=[('scale', StandardScaler()),\n",
       "                                       ('sampling',\n",
       "                                        SMOTE(sampling_strategy='minority')),\n",
       "                                       ('logreg',\n",
       "                                        LogisticRegression(max_iter=1000,\n",
       "                                                           solver='saga'))]),\n",
       "             n_jobs=-1,\n",
       "             param_grid={'logreg__C': [1], 'logreg__penalty': ['l2', 'l1'],\n",
       "                         'sampling__k_neighbors': [5, 10]},\n",
       "             scoring='roc_auc')"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "cfbe1b24",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'logreg__C': 1, 'logreg__penalty': 'l2', 'sampling__k_neighbors': 10}"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "ea8ea2ed",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8301235839340886, 0.7226018195078884)"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.score(X_train, y_train), grid.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "6855c70f",
   "metadata": {},
   "source": [
    "Accuracy for train set is 0.831, while test set is 0.726. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "e448651b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# getting feature importance \n",
    "lr_coef = grid.best_estimator_.named_steps['logreg'].coef_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "7a356aba",
   "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>variable</th>\n",
       "      <th>coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>bun_min</td>\n",
       "      <td>1.740401</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>wbc_max</td>\n",
       "      <td>1.071415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>bicarbonate_min</td>\n",
       "      <td>0.622655</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>temperature_mean</td>\n",
       "      <td>0.556780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hematocrit_min</td>\n",
       "      <td>0.553349</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>chloride_min</td>\n",
       "      <td>0.525415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>hemoglobin_max</td>\n",
       "      <td>0.453564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>calcium_min</td>\n",
       "      <td>0.391762</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>creatinine_max</td>\n",
       "      <td>0.391611</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>ptt_max</td>\n",
       "      <td>0.386170</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>pt_max</td>\n",
       "      <td>0.275772</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>weight</td>\n",
       "      <td>0.143297</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>chloride_max</td>\n",
       "      <td>0.120558</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>aniongap_min</td>\n",
       "      <td>0.102348</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>platelets_max</td>\n",
       "      <td>0.088927</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>admission_age</td>\n",
       "      <td>0.079800</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>inr_max</td>\n",
       "      <td>0.051411</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>spo2_mean</td>\n",
       "      <td>0.042039</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>heart_rate_mean</td>\n",
       "      <td>-0.008651</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>glucose_min</td>\n",
       "      <td>-0.059023</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>glucose_mean</td>\n",
       "      <td>-0.068734</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>platelets_min</td>\n",
       "      <td>-0.077091</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>mbp_mean</td>\n",
       "      <td>-0.080034</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>glucose_max</td>\n",
       "      <td>-0.091498</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>gcs_min</td>\n",
       "      <td>-0.093727</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>potassium_min</td>\n",
       "      <td>-0.124327</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>aniongap_max</td>\n",
       "      <td>-0.130542</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>aki_stage</td>\n",
       "      <td>-0.176617</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>potassium_max</td>\n",
       "      <td>-0.191336</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>sodium_min</td>\n",
       "      <td>-0.227896</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>calcium_max</td>\n",
       "      <td>-0.254520</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>sodium_max</td>\n",
       "      <td>-0.273513</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>hematocrit_max</td>\n",
       "      <td>-0.299659</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>resp_rate_mean</td>\n",
       "      <td>-0.348373</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>bicarbonate_max</td>\n",
       "      <td>-0.470534</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hemoglobin_min</td>\n",
       "      <td>-0.715982</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>wbc_min</td>\n",
       "      <td>-0.851116</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>bun_max</td>\n",
       "      <td>-1.122809</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>creatinine_min</td>\n",
       "      <td>-1.151577</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            variable      coef\n",
       "15           bun_min  1.740401\n",
       "10           wbc_max  1.071415\n",
       "13   bicarbonate_min  0.622655\n",
       "35  temperature_mean  0.556780\n",
       "3     hematocrit_min  0.553349\n",
       "19      chloride_min  0.525415\n",
       "6     hemoglobin_max  0.453564\n",
       "17       calcium_min  0.391762\n",
       "22    creatinine_max  0.391611\n",
       "31           ptt_max  0.386170\n",
       "30            pt_max  0.275772\n",
       "0             weight  0.143297\n",
       "20      chloride_max  0.120558\n",
       "11      aniongap_min  0.102348\n",
       "8      platelets_max  0.088927\n",
       "2      admission_age  0.079800\n",
       "29           inr_max  0.051411\n",
       "36         spo2_mean  0.042039\n",
       "32   heart_rate_mean -0.008651\n",
       "23       glucose_min -0.059023\n",
       "37      glucose_mean -0.068734\n",
       "7      platelets_min -0.077091\n",
       "33          mbp_mean -0.080034\n",
       "24       glucose_max -0.091498\n",
       "38           gcs_min -0.093727\n",
       "27     potassium_min -0.124327\n",
       "12      aniongap_max -0.130542\n",
       "1          aki_stage -0.176617\n",
       "28     potassium_max -0.191336\n",
       "25        sodium_min -0.227896\n",
       "18       calcium_max -0.254520\n",
       "26        sodium_max -0.273513\n",
       "4     hematocrit_max -0.299659\n",
       "34    resp_rate_mean -0.348373\n",
       "14   bicarbonate_max -0.470534\n",
       "5     hemoglobin_min -0.715982\n",
       "9            wbc_min -0.851116\n",
       "16           bun_max -1.122809\n",
       "21    creatinine_min -1.151577"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_tot = pd.DataFrame(list(zip(X.columns,lr_coef[0])))\n",
    "coef_tot.columns = ['variable','coef']\n",
    "coef_tot.sort_values(by='coef',ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "20593a99",
   "metadata": {},
   "source": [
    "**top 5 features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "c42c241a",
   "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>variable</th>\n",
       "      <th>coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>bun_min</td>\n",
       "      <td>1.740401</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>wbc_max</td>\n",
       "      <td>1.071415</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>bicarbonate_min</td>\n",
       "      <td>0.622655</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>temperature_mean</td>\n",
       "      <td>0.556780</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>hematocrit_min</td>\n",
       "      <td>0.553349</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "            variable      coef\n",
       "15           bun_min  1.740401\n",
       "10           wbc_max  1.071415\n",
       "13   bicarbonate_min  0.622655\n",
       "35  temperature_mean  0.556780\n",
       "3     hematocrit_min  0.553349"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_tot.sort_values(by='coef',ascending=False).head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "64e4da59",
   "metadata": {},
   "source": [
    "There is 18x higher likelihood to receive delayed rrt with one unit increase in bun_min, compared to features such as 'platelets_max', 'inr_max', 'glucose_min', 'chloride_max', 'aniongap_min', 'heart_rate_mean', 'glucose_max', 'glucose_mean',\t'mbp_mean', 'platelets_min'."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ff97035",
   "metadata": {},
   "source": [
    "**negative features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "70a3a200",
   "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>variable</th>\n",
       "      <th>coef</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>creatinine_min</td>\n",
       "      <td>-1.151577</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>bun_max</td>\n",
       "      <td>-1.122809</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>wbc_min</td>\n",
       "      <td>-0.851116</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>hemoglobin_min</td>\n",
       "      <td>-0.715982</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>bicarbonate_max</td>\n",
       "      <td>-0.470534</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "           variable      coef\n",
       "21   creatinine_min -1.151577\n",
       "16          bun_max -1.122809\n",
       "9           wbc_min -0.851116\n",
       "5    hemoglobin_min -0.715982\n",
       "14  bicarbonate_max -0.470534"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "coef_tot.sort_values(by='coef').head()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "67099af4",
   "metadata": {},
   "source": [
    "There is 12x higher likelihood to NOT receive delayed rrt with one unit increase in bun_max, compared to features such as 'platelets_max', 'inr_max', 'glucose_min', 'chloride_max', 'aniongap_min', 'heart_rate_mean', 'glucose_max', 'glucose_mean',\t'mbp_mean', 'platelets_min'."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "49b57224",
   "metadata": {},
   "source": [
    "## Feature extraction"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "65b77570",
   "metadata": {},
   "source": [
    "For feature extraction, we will use PCA to identify and keep those potential important relationships, before performing the predictions/models. Due to the number of features (and high-dimensional manifold) in our dataset, PCA was utilised to perform an unsupervised dimensionality reduction on the training dataset. In such a way, PCA will provide the best linear approximations before modeling and prediction was performed ([source](https://arxiv.org/ftp/arxiv/papers/1403/1403.1949.pdf)). GridSearchCV is used to set the dimensionality of the PCA. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "220b52c1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "aniongap_max potassium_max       0.291583\n",
       "aniongap_max creatinine_max      0.290806\n",
       "aniongap_max mbp_mean            0.274197\n",
       "creatinine_max potassium_max     0.273823\n",
       "aniongap_max dbp_mean            0.262543\n",
       "aniongap_min potassium_max       0.260355\n",
       "calcium_max creatinine_max       0.259650\n",
       "hematocrit_max creatinine_max    0.259116\n",
       "aniongap_max sbp_mean            0.259088\n",
       "aniongap_min creatinine_max      0.258068\n",
       "hemoglobin_max creatinine_max    0.257905\n",
       "creatinine_max gcs_min           0.257502\n",
       "creatinine_max resp_rate_mean    0.256791\n",
       "aniongap_max^2                   0.255885\n",
       "aniongap_max                     0.254689\n",
       "aniongap_max calcium_max         0.254303\n",
       "aniongap_max spo2_mean           0.250221\n",
       "aniongap_max sodium_max          0.249964\n",
       "aniongap_max temperature_mean    0.249200\n",
       "aniongap_max gcs_min             0.248238\n",
       "dtype: float64"
      ]
     },
     "execution_count": 39,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# including polynomial features\n",
    "abs(X_poly_corrs).sort_values(ascending=False).head(20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "aecd6f6b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Adding top abs 20 interaction features into dataset\n",
    "# aniongap_max^2 is not added, because little diff between aniongap_max (which is already in dataset)\n",
    "\n",
    "aki_data['aniongap_max*potassium_max'] = aki_data['aniongap_max'] * aki_data['potassium_max']\n",
    "aki_data['aniongap_max*creatinine_max'] = aki_data['aniongap_max'] * aki_data['creatinine_max']\n",
    "aki_data['aniongap_max*mbp_mean'] = aki_data['aniongap_max'] * aki_data['mbp_mean']\n",
    "aki_data['creatinine_max*potassium_max'] = aki_data['creatinine_max'] * aki_data['potassium_max']\n",
    "aki_data['aniongap_min*potassium_max'] = aki_data['aniongap_min'] * aki_data['potassium_max']\n",
    "aki_data['calcium_max*creatinine_max'] = aki_data['calcium_max'] * aki_data['creatinine_max']\n",
    "aki_data['hematocrit_max*creatinine_max'] = aki_data['hematocrit_max'] * aki_data['creatinine_max']\n",
    "aki_data['aniongap_min*creatinine_max'] = aki_data['aniongap_min'] * aki_data['creatinine_max']\n",
    "aki_data['hemoglobin_max*creatinine_max'] = aki_data['hemoglobin_max'] * aki_data['creatinine_max']\n",
    "aki_data['creatinine_max*gcs_min'] = aki_data['creatinine_max'] * aki_data['gcs_min']\n",
    "aki_data['creatinine_max*resp_rate_mean'] = aki_data['creatinine_max'] = aki_data['resp_rate_mean']\n",
    "aki_data['aniongap_max*calcium_max'] = aki_data['aniongap_max'] * aki_data['calcium_max']\n",
    "aki_data['aniongap_max*spo2_mean'] = aki_data['aniongap_max'] * aki_data['spo2_mean']\n",
    "aki_data['aniongap_max*sodium_max'] = aki_data['aniongap_max'] * aki_data['sodium_max']\n",
    "aki_data['aniongap_max*temperature_mean'] = aki_data['aniongap_max'] * aki_data['temperature_mean']\n",
    "aki_data['aniongap_max*gcs_min'] = aki_data['aniongap_max'] * aki_data['gcs_min']\n",
    "# aki_data['chloride_min*temperature_mean'] = aki_data['chloride_min'] * aki_data['temperature_min']\n",
    "# aki_data['bicarbonate_min*chloride_max'] = aki_data['bicarbonate_min'] * aki_data['chloride_max']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "b059bff2",
   "metadata": {},
   "outputs": [],
   "source": [
    "X = aki_data.drop(columns=['delay_rrt'])\n",
    "y = aki_data[['delay_rrt']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "7cafb96c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(1778, 55)"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# ? + 18 = 55\n",
    "X.shape "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "b89f9468",
   "metadata": {},
   "outputs": [],
   "source": [
    "# split\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.33, random_state=42, stratify=y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "6a9401ec",
   "metadata": {},
   "outputs": [],
   "source": [
    "# scaling\n",
    "ss_pca = StandardScaler()\n",
    "\n",
    "X_train_sc = ss_pca.fit_transform(X_train)\n",
    "X_test_sc = ss_pca.transform(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "id": "8d78ffed",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "PCA(n_components=55)"
      ]
     },
     "execution_count": 45,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Instantiate PCA.\n",
    "# up to n_components were tested\n",
    "pca = PCA(n_components = 55)\n",
    "\n",
    "# Fit PCA on the training data.\n",
    "pca.fit(X_train_sc)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f71fcfc9",
   "metadata": {},
   "source": [
    "In the below pipeline, PCA will perform unsupervised dimensionality reduction on the training data before upsampling it. This is so that upsampling will only be performed on the reduced dimensional manifold and not on the entire dataset. The models for the prediction will be done last. In the below pipeline, the logistic regression model was utilised. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "id": "dd39438b",
   "metadata": {},
   "outputs": [],
   "source": [
    "pipe = Pipeline([\n",
    "        ('scale', StandardScaler()),\n",
    "        ('pca', PCA()),\n",
    "        ('sampling', SMOTE(sampling_strategy='minority')),\n",
    "        ('logreg', LogisticRegression(max_iter=1_000, solver='saga'))\n",
    "    ])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "e5a93de3",
   "metadata": {},
   "outputs": [],
   "source": [
    "pipe_params = {\n",
    "    'pca__n_components': [5,10,15,20,25,30,35,40,45,50,55],  \n",
    "} "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "385dec3e",
   "metadata": {},
   "outputs": [],
   "source": [
    "grid = GridSearchCV(pipe, pipe_params, scoring='roc_auc', n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "8dac0886",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "GridSearchCV(estimator=Pipeline(steps=[('scale', StandardScaler()),\n",
       "                                       ('pca', PCA()),\n",
       "                                       ('sampling',\n",
       "                                        SMOTE(sampling_strategy='minority')),\n",
       "                                       ('logreg',\n",
       "                                        LogisticRegression(max_iter=1000,\n",
       "                                                           solver='saga'))]),\n",
       "             n_jobs=-1,\n",
       "             param_grid={'pca__n_components': [5, 10, 15, 20, 25, 30, 35, 40,\n",
       "                                               45, 50, 55]},\n",
       "             scoring='roc_auc')"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "af15f3bf",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# https://scikit-learn.org/stable/auto_examples/compose/plot_digits_pipe.html#sphx-glr-auto-examples-compose-plot-digits-pipe-py\n",
    "# Plot search for best combination of PCA n_components and logistic regression accuracy\n",
    "# pca.fit(X_train_sc)\n",
    "fig, (ax0, ax1) = plt.subplots(nrows=2, sharex=True, figsize=(6, 6))\n",
    "ax0.plot(np.arange(1, pca.n_components_ + 1),\n",
    "         pca.explained_variance_ratio_.cumsum(), '+', linewidth=10)\n",
    "ax0.set_ylabel('Cumulative PCA explained variance ratio')\n",
    "\n",
    "results = pd.DataFrame(grid.cv_results_)\n",
    "components_col = 'param_pca__n_components'\n",
    "best_clfs = results.groupby(components_col).apply(\n",
    "    lambda g: g.nlargest(1, 'mean_test_score'))\n",
    "\n",
    "best_clfs.plot(x=components_col, y='mean_test_score', yerr='std_test_score',\n",
    "               legend=False, ax=ax1)\n",
    "ax1.set_ylabel('Classification accuracy (val)')\n",
    "ax1.set_xlabel('n_components');\n",
    "plt.suptitle('Dimensional reduction based on PCA n_components');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "9586d9d2",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Cumulative explained variance (for n_components=20): 0.9317110347308232\n"
     ]
    }
   ],
   "source": [
    "print(f'Cumulative explained variance (for n_components=20): {pca.explained_variance_ratio_.cumsum()[20]}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "46f4aa5b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'pca__n_components': 45}"
      ]
     },
     "execution_count": 52,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "223ee140",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.8301048590955904, 0.7202410656020881)"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "grid.score(X_train, y_train), grid.score(X_test, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "38f3e7fd",
   "metadata": {},
   "source": [
    "GridSearchCV chose n_components=45, however, there was overfitting at n_components=45. Since PCA utilises n_components, the decision is to use n_components=20, to minimise overfitting, and we would retain approximately 93% of the variability in the data.\n",
    "<br>\n",
    "<br> The curse is after the dataset is PCA-transformed, there is no correlation between any features. In addition, the new principal components are not interpretable. So, to dig into the feature importance, one will have to convert the PCA-transformed data back to the original features."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "41531586",
   "metadata": {},
   "source": [
    "### Models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "bcb07cf0",
   "metadata": {},
   "outputs": [],
   "source": [
    "# Instantiate models\n",
    "models = {'lr': LogisticRegression(max_iter=1000, solver='saga'),\n",
    "        'dt': DecisionTreeClassifier(),\n",
    "        'rf': RandomForestClassifier(),\n",
    "        'et': ExtraTreesClassifier(),\n",
    "        'svc': SVC(probability=True),\n",
    "        } "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "id": "80a0d8e1",
   "metadata": {},
   "outputs": [],
   "source": [
    "def run_models(model, model_params):\n",
    "    pipe = Pipeline([\n",
    "            ('scale', StandardScaler()),\n",
    "            ('pca', PCA(n_components = 30)),\n",
    "            ('sampling', SMOTE(sampling_strategy = 'minority')),\n",
    "            (model, models[model])\n",
    "            ])\n",
    "    \n",
    "    grid = GridSearchCV(pipe, param_grid=model_params, cv=3, scoring='roc_auc', n_jobs=-1)\n",
    "    grid.fit(X_train, y_train)\n",
    "\n",
    "    # metrics\n",
    "    train_score = grid.score(X_train, y_train)\n",
    "    test_score = grid.score(X_test, y_test)\n",
    "    preds = grid.predict(X_test)\n",
    "    tn, fp, fn, tp = confusion_matrix(y_test, preds).ravel()\n",
    "  \n",
    "    # View confusion matrix\n",
    "    plot_confusion_matrix(grid, X_test, y_test, cmap='Blues', values_format='d');\n",
    "    \n",
    "    # Calculate the sensitivity/ recall\n",
    "    sens = tp / (tp + fn)\n",
    "    \n",
    "    # Calculate the specificity\n",
    "    spec = tn / (tn + fp) \n",
    "    \n",
    "    # print results\n",
    "    print(f'Best params: {grid.best_params_}')\n",
    "    print(f'Training score: {round(train_score, 4)}')\n",
    "    print(f'Test score: {round(test_score, 4)}')\n",
    "    print(f'Recall: {round(sens, 4)}')\n",
    "    print(f'Specificity: {round(spec, 4)}')\n",
    "    return grid"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "82bbaa2d",
   "metadata": {},
   "source": [
    "**Logistic Regression**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "fd638152",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params: {'lr__C': 0.1, 'lr__penalty': 'l1', 'sampling__k_neighbors': 10}\n",
      "Training score: 0.8032\n",
      "Test score: 0.7273\n",
      "Recall: 0.7406\n",
      "Specificity: 0.5872\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# logistic regression\n",
    "# grid_params = {\n",
    "#     'sampling__k_neighbors': [10,15,20],  \n",
    "#     'lr__penalty': ['l1', 'l2',  'elasticnet'],   \n",
    "#     'lr__C': [0.1, 1, 10]  \n",
    "# } \n",
    "\n",
    "best_params = {\n",
    "    'sampling__k_neighbors': [10],  \n",
    "    'lr__penalty': ['l1'],   \n",
    "    'lr__C': [0.1]  \n",
    "# }\n",
    "\n",
    "lr = run_models('lr', best_params)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f466eac7",
   "metadata": {},
   "source": [
    "**Decision Tree Classifier**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "6a65e132",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params: {'dt__max_depth': 70, 'dt__min_samples_leaf': 90, 'dt__min_samples_split': 90, 'sampling__k_neighbors': 60}\n",
      "Training score: 0.7909\n",
      "Test score: 0.6947\n",
      "Recall: 0.636\n",
      "Specificity: 0.6422\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# decision tree classifier\n",
    "# grid_params = {\n",
    "#     'sampling__k_neighbors': [40, 50, 60],  \n",
    "#     'dt__max_depth': [70, 80, 90],\n",
    "#     'dt__min_samples_split': [70, 80, 90],\n",
    "#     'dt__min_samples_leaf': [70, 80, 90],\n",
    "# } \n",
    "\n",
    "best_params = {\n",
    "    'sampling__k_neighbors': [60],  \n",
    "    'dt__max_depth': [70],\n",
    "    'dt__min_samples_split': [90],\n",
    "    'dt__min_samples_leaf': [90],\n",
    "} \n",
    "\n",
    "dt = run_models('dt', best_params)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "80c83052",
   "metadata": {},
   "source": [
    "**Random forest classifier**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "e3a7ee79",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params: {'rf__max_depth': 10, 'rf__min_samples_leaf': 50, 'rf__min_samples_split': 20, 'rf__n_estimators': 20, 'sampling__k_neighbors': 30}\n",
      "Training score: 0.8638\n",
      "Test score: 0.6855\n",
      "Recall: 0.7427\n",
      "Specificity: 0.4679\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# random forest classifier\n",
    "# grid_params = {\n",
    "#     'sampling__k_neighbors': [70, 80, 90],\n",
    "#     'rf__n_estimators': [70, 80, 90],\n",
    "#     'rf__max_depth': [50, 80, 90],\n",
    "#     'rf__min_samples_split': [70, 80, 90],\n",
    "#     'rf__min_samples_leaf': [70, 80, 90],\n",
    "# } \n",
    "\n",
    "best_params = {\n",
    "    'sampling__k_neighbors': [30], \n",
    "    'rf__n_estimators': [20],\n",
    "    'rf__max_depth': [10],\n",
    "    'rf__min_samples_split': [20],\n",
    "    'rf__min_samples_leaf': [50],\n",
    "} \n",
    "\n",
    "rf = run_models('rf', best_params)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "86b77586",
   "metadata": {},
   "source": [
    "**Extra trees classifier**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "f99da4fe",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params: {'et__max_depth': 4, 'et__min_samples_leaf': 10, 'et__min_samples_split': 15, 'et__n_estimators': 35, 'sampling__k_neighbors': 15}\n",
      "Training score: 0.857\n",
      "Test score: 0.7277\n",
      "Recall: 0.7322\n",
      "Specificity: 0.578\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# extra trees classifier\n",
    "# grid_params = {\n",
    "#     'sampling__k_neighbors': [15, 20, 25],\n",
    "#     'et__n_estimators': [30, 35],\n",
    "#     'et__max_depth': [4, 5, 6],\n",
    "#     'et__min_samples_split': [10, 15],\n",
    "#     'et__min_samples_leaf': [10, 15],\n",
    "# } \n",
    "\n",
    "best_params = {\n",
    "    'sampling__k_neighbors': [20],\n",
    "    'et__n_estimators': [30],  \n",
    "    'et__max_depth': [4],  \n",
    "    'et__min_samples_split': [15],\n",
    "    'et__min_samples_leaf': [10],\n",
    "}\n",
    "\n",
    "et = run_models('et', best_params)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ad9b4678",
   "metadata": {},
   "source": [
    "**Feature importance**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "26b952a1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1440x1080 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# https://stackoverflow.com/questions/50796024/feature-variable-importance-after-a-pca-analysis\n",
    "# PCA-transform data back to orignal features\n",
    "num_plot = 20\n",
    "\n",
    "pipe = Pipeline([\n",
    "            ('scale', StandardScaler()),\n",
    "            ('pca', PCA(n_components = num_plot)),\n",
    "            ('sampling', SMOTE(sampling_strategy='minority', k_neighbors=20)),\n",
    "            ('rf', ExtraTreesClassifier(n_estimators=30, max_depth=6, min_samples_split = 10, min_samples_leaf = 15)),\n",
    "            ])\n",
    "\n",
    "pipe.fit(X_train, y_train)\n",
    "\n",
    "model = pipe.steps[1][1]\n",
    "n_pcs = model.components_.shape[0]\n",
    "\n",
    "# get index for more important feature from each n_component\n",
    "feature_names = X_train.columns\n",
    "most_important = [np.abs(model.components_[i]).argmax() for i in range(n_pcs)]\n",
    "most_important_names = [feature_names[most_important[i]] for i in range(n_pcs)]\n",
    "\n",
    "# combine for plotting later\n",
    "zipped_feats = zip(most_important_names, pipe.steps[3][1].feature_importances_)\n",
    "zipped_feats = sorted(zipped_feats, key=lambda x: x[1], reverse=True)\n",
    "features, importances = zip(*zipped_feats)\n",
    "\n",
    "# n_components = 20\n",
    "top_features = features[:num_plot]\n",
    "top_importances = importances[:num_plot]\n",
    "plt.figure(figsize=(20,15))\n",
    "plt.barh(range(len(top_importances)), top_importances, align='center')\n",
    "plt.yticks(range(len(top_importances)), top_features)\n",
    "plt.title('Feature Importances',fontsize=20)\n",
    "plt.xlabel('Gini Importance',fontsize=15)\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "37eed970",
   "metadata": {},
   "source": [
    "**Support Vector Classification**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "58f35d9f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Best params: {'sampling__k_neighbors': 30, 'svc__C': 10, 'svc__gamma': 'scale', 'svc__kernel': 'linear'}\n",
      "Training score: 0.8063\n",
      "Test score: 0.7324\n",
      "Recall: 0.7176\n",
      "Specificity: 0.6055\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Support Vector Classification\n",
    "# grid_params = {\n",
    "#     'sampling__k_neighbors': [25, 30],     \n",
    "#     'svc__kernel': ['linear', 'sigmoid'],   \n",
    "#     'svc__gamma' : ['scale', 'auto'], \n",
    "#     'svc__C': [1 ,5 ,10] \n",
    "# } \n",
    "\n",
    "best_params = {\n",
    "    'sampling__k_neighbors': [30],     \n",
    "    'svc__kernel': ['linear'],    \n",
    "    'svc__gamma' : ['scale'],\n",
    "    'svc__C': [10] \n",
    "} \n",
    "\n",
    "svc = run_models('svc', best_params)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d2f53b91",
   "metadata": {},
   "source": [
    "**Feature importance**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "dec7737a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('aniongap_max*creatinine_max',)\n",
      "(array([-0.26403962, -0.08274747, -0.12295176,  0.0170741 , -0.07941787,\n",
      "       -0.01720758,  0.23581009,  0.24812305, -0.27333569,  0.16784321,\n",
      "       -0.14063603, -0.05085818,  0.0607359 , -0.12750844, -0.53725808,\n",
      "       -0.2241743 ,  0.0583987 , -0.05377906, -0.04928964, -0.00105487,\n",
      "        0.32070493, -0.02713314, -0.18489336, -0.24918143, -0.08455037,\n",
      "        0.12466718, -0.17747838, -0.52683501, -0.21476526, -0.28266959]),)\n"
     ]
    }
   ],
   "source": [
    "# https://stackoverflow.com/questions/50796024/feature-variable-importance-after-a-pca-analysis\n",
    "# PCA-transform data back to orignal features\n",
    "num_plot = 30\n",
    "\n",
    "pipe = Pipeline([\n",
    "            ('scale', StandardScaler()),\n",
    "            ('pca', PCA(n_components = num_plot)),\n",
    "            ('sampling', SMOTE(sampling_strategy='minority',k_neighbors =30)),\n",
    "            ('svc', SVC(C=10, gamma='auto', kernel='linear',probability=True)),\n",
    "            ])\n",
    "\n",
    "pipe.fit(X_train, y_train)\n",
    "\n",
    "model = pipe.steps[1][1]\n",
    "n_pcs = model.components_.shape[0]\n",
    "\n",
    "# get index for more important feature from each n_component\n",
    "feature_names = X_train.columns\n",
    "most_important = [np.abs(model.components_[i]).argmax() for i in range(n_pcs)]\n",
    "most_important_names = [feature_names[most_important[i]] for i in range(n_pcs)]\n",
    "\n",
    "# combine for plotting later\n",
    "zipped_feats = zip(most_important_names, pipe.steps[3][1].coef_)\n",
    "zipped_feats = sorted(zipped_feats, key=lambda x: x[1], reverse=True)\n",
    "features, importances = zip(*zipped_feats)\n",
    "\n",
    "# n_components = 30\n",
    "top_features = features[:num_plot]\n",
    "top_importances = importances[:num_plot]\n",
    "print(top_features)\n",
    "print(top_importances)\n",
    "# plt.figure(figsize=(10,7))\n",
    "# plt.barh(range(len(top_importances)), top_importances, align='center')\n",
    "# plt.yticks(range(len(top_importances)), top_features)\n",
    "# plt.title('Feature Importances')\n",
    "# plt.xlabel('Coefficients')\n",
    "# plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9632ddc8",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b3e11b62",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "id": "31b9e55e",
   "metadata": {},
   "source": [
    "## Plot ROC-AUC "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "2d33eff1",
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_all_auc(models):\n",
    "    fig, ax = plt.subplots(figsize=(10,10))\n",
    "    axes = {}\n",
    "    for i, m in enumerate(models.keys()):\n",
    "        axes[f'ax{i}'] = plot_roc_curve(m, X_test, y_test, ax=ax, name=models[m])\n",
    "        \n",
    "    plt.plot([0, 1], [0, 1], color='k', linestyle='--')\n",
    "    plt.title('ROC-AUC curve',fontsize=20)\n",
    "    plt.xlabel('False Positive Rate (FPR)',fontsize=16)\n",
    "    plt.ylabel('True Positive Rate (TPR)',fontsize=16)\n",
    "    plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "bcede07b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "models_built = {\n",
    "    lr : 'LogisticRegression', \n",
    "    dt : 'DecisionTreeClassifier',\n",
    "    rf : 'RandomForestClassifier',\n",
    "    et : 'ExtraTreesClassifier',\n",
    "    svc : 'SVC',\n",
    "} \n",
    "\n",
    "plot_all_auc(models_built)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "46c43b5d",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.8.8"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {
    "height": "calc(100% - 180px)",
    "left": "10px",
    "top": "150px",
    "width": "384px"
   },
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}