Diabetes_Hospitalizations.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Diabetes Hospitalizations\n",
    "\n",
    "#### Aren Carpenter & Jean-Paul Ventura\n",
    "\n",
    "FS Data Science Cohort 062220"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import Libraries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "pd.set_option('display.max_columns', 120)\n",
    "import scipy.stats as scs\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "from imblearn.under_sampling import TomekLinks\n",
    "from imblearn.over_sampling import SMOTE\n",
    "\n",
    "from sklearn.model_selection import train_test_split, GridSearchCV\n",
    "from sklearn.preprocessing import PolynomialFeatures, StandardScaler\n",
    "from sklearn.metrics import make_scorer, precision_score, recall_score, accuracy_score, f1_score\n",
    "from sklearn.linear_model import LogisticRegressionCV\n",
    "from sklearn.feature_selection import RFECV\n",
    "from sklearn.dummy import DummyClassifier\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "\n",
    "import warnings\n",
    "warnings.filterwarnings('ignore')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Location of Dataset:\n",
    "\n",
    "https://data.world/uci/diabetes-130-us-hospitals-for-years-1999-2008\n",
    "\n",
    "### Article for which data was collected originally:\n",
    "https://www.hindawi.com/journals/bmri/2014/781670/"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import Data and Cleaning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_csv('https://query.data.world/s/fzhdybgova7pqh6amwfzrnhumdc26t')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Need to remove time-series elements from when the same patient visited several times. Kept the first entry and removed all others."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop_duplicates(subset='patient_nbr', inplace=True) # 101,766 to 71,518 observations"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Weight, Payer Code, and Medical Specialty were missing >50% of values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['encounter_id','patient_nbr','weight', 'payer_code', 'medical_specialty'], axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Dropping missing values from Race column."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[df.race != '?'] # about 1,000 obs\n",
    "df = df[df.gender != 'Unknown/Invalid'] # 1 obs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Changed target variable to 0-2 for classes (No readmit, less than 30 days, more than 30 days)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.readmitted.replace({'NO': 0, '<30': 1, '>30': 2}, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Parsing the ICD-9 medical billing codes, in order to grab just 250.xx coding diabetes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[pd.to_numeric(df['diag_1'], errors='coerce').notnull()] # Select non-null values after changing values to\n",
    "df = df[pd.to_numeric(df['diag_2'], errors='coerce').notnull()] # numeric (removing V27/V57 entries and ?)\n",
    "df = df[pd.to_numeric(df['diag_3'], errors='coerce').notnull()] #\n",
    "\n",
    "df.diag_1 = df.diag_1.astype('float64') # Can now recast from Object to Float64 dtypes (want float to preserve\n",
    "df.diag_2 = df.diag_2.astype('float64') # ICD-9 decimals as part of the billing code)\n",
    "df.diag_3 = df.diag_3.astype('float64') #"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## EDA"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We wanted to explore some demographic features as they relate to readmission rates. We started with the actual breakdown of readmission classes. There is some class imbalance, so we will try SMOTE/Tomek later to try to improve our predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "x = ['Not Readmitted', 'Within 30 Days', 'After 30 Days']\n",
    "y = [len(df[df.readmitted == 0]), len(df[df.readmitted == 1]), len(df[df.readmitted == 2])]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAgQAAAHjCAYAAABcqwcxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3deZxddX3/8dfbsIgboESkBAlqqoKtESJStZWqhYBVsNUKdYmWNl2gLnVD2xoQafVnFYsLFSUNWDTFrUQbjRGxuLAFCEtEJAKVCEJsEEEUCnx+f5zvlMtwJzMTJneSyev5eJzHPed7vufc77n3zL3v+z3LpKqQJElbtodMdgMkSdLkMxBIkiQDgSRJMhBIkiQMBJIkCQOBJEnCQCBt0ZJ8OcmiB7mOmUkqyZwJaM9rk9z+YNcjafwMBNIkSrKofZlWkruT/CjJSUl2nOy2jcP1wC7AyglY178DT5iA9axXkmN6Xvd7k9yQ5PQku41zPfu3dey0sdoqDYqBQJp8X6f7Qp0J/CnwYuBjk9mg8aiqe6rqJ1V19wSs65dVdfNEtGsMrqJ73WcArwB+AzhjQM8tbXIMBNLku7N9oa6pqq/R/Uo+oLdCku2TnJzk5iS3Jfmv3i76JI9J8pkka5L8MsmqJK8bto6HtR6J25PclOSdwxuS5Lok72r1bktyfZJXJNkhyeK27NVJDuhZ5n6HDJJsneTE9qv7zraO9/bU/4Mkl7V2rmvbsnOb94BDBkn+PMnqJHe1xz8bNr+SzE/y2SS/SHJNkleN4XW/u73uN1TVt4BPAPsleVTPul+V5ML2WtzcnmPXoe0Gzm5V17Z2LGrzkuRtSX7YtvPyMbZJmjQGAmkTkuQJwFzgf3vKAvwnsCvw+8AzgHOAbyTZpVV7KHBxm78X8M/Ax5O8oGf1/wT8HvCHwAvaen6nTzPeCFwA7E33i/lU4NPAUmB2e+5/S/LQETbj9cBLgcOAWXS/vq9q2/I4YHFb51Pb839qPa/HS4GPAB8Cnta262NJXjys6ruAM4Gn0wWqhUl2H2m9fZ7nccAfAPe0Ycg2wIK23t8HdgI+0+ZdT/daQvea7wK8oU2/BzgCOBLYE/hHuvfjRWNtkzRwVeXg4DBJA7AIuBu4HfglUG14U0+d57f52w1bdiXwtvWsezHwyTb+COBO4JU98x8B/AxY1FN2HfCZYXUKOLGnbGYrmzPC9InAWUD6tGnvVnf3Edr8WuD2nunvAAv7vGbf7pku4B97prcC7gBetZ7X5hi6L/7bW92h1/2fR3m/ntLqzWjT+7fpnXrqPLy9l789bNkPAUsne59zcBhp2Gr0yCBpIzsHmA9sB/wZ8ES6L9Uh+wAPo+uW7l3uoa0uSaYBR9P9Gt8V2Jbu1+03W90ntulzhxauqtuTXN6nPZcNq3MH0Fvvpvb42BG2ZxGwHPhBkq/R9Sx8paruBS6lO2fiijbv68DnqmrtCOt6KrBwWNm3gZesp813J1m7nvYN+SFwMN1rdQjdr/37HUZJsjddD8Fs4NHA0BvweGDNCOvdk+69+WqS3v8etzVd4JI2SQYCafLdUVWr2/jrk5wN/D3dr1joDu3dBPx2n2V/3h7fAryZrsv6crpfvv/AfV+KeeCiI/rfYdM1rGzoS67vIcequrgdX59L17txKnBpkt+rqnva+Qf70Z0ncQTwj0meV1WXjtCefv+SdXhZvzaPdkj0rp7XfVWSWcBH6XopSPJwYBldaHk1cDPdIYNv0YWrkQw974uBH43STmmT4TkE0qbnWODtSX6tTV8M7AzcW1Wrhw1DZ+Q/F/hSVX2qqlbS/fr99Z51rqb7MtpvqKB94T1tY2xAVd1WVZ+tqr8EXkQXDJ7U5lVVnVtVxwLPBG6g69no50q6bev1XOB7G6HZxwGvTLJPm34KXQB4Z1WdU1Xf54G9Dne1x2k9Zd+jOzyze5/36783QrulCWEPgbSJqapvJlkF/B3wV3S/UL8DnJnkbcD3gcfR/QL/enVnyP8AeEWS5wI/Bf4a2AO4pK3z9iSnAO9r3ek30J2IN40JluRvgBvpznH4X+CP6Xoy1iTZD3gh3S/vm+hObNyNkb/g3w98NslFwNfaNr+S7gTACVVV1yRZQhcMDqb7dX8ncFSSj9Idvjhu2GL/Tdcb8aIkXwJ+WVW3Jfkn4J/aCaHn0J2LsR9dqDt5otsuTQR7CKRN0weBI5LsXlVF9wX1DbpL466iO/v/yXRf7NCd1X4B8BW6L6BfAKcPW+db6C6T+2J7vKLVnWi3AW9t7bmY7vj7QVV1B3Ar8Bzgy8DVwAeA46rq3/qtqKr+gy7cvIkuNLwB+Kuq+tJGaDetPQcleXY7r2EecGh77gXA3wxr349b+fF0AecjbdbQIZ+3AKvozqn4Q+DajdRu6UFL91kjSZK2ZPYQSJIkA4EkSTIQSJIkDASSJAkDgSRJYgu/D8FOO+1UM2fOnOxmSJI0EBdddNFPq2p6v3lbdCCYOXMmK1asmOxmSJI0EElGvFumhwwkSZKBQJIkGQgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSsNVkN2AqufXYYye7CZpA2y9YMNlNkKSBsYdAkiQZCCRJkoFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZLEgAJBkocmuSDJpUlWJTm2lS9Kcm2SlW2Y3cqT5MQkq5NclmTvnnXNS3J1G+b1lO+T5PK2zIlJMohtkyRpKthqQM9zJ/D8qro9ydbAt5N8pc17a1V9blj9g4BZbXgWcBLwrCSPBhYAc4ACLkqypKpuaXXmA+cBS4G5wFeQJEmjGkgPQXVub5Nbt6HWs8ghwGltufOAHZLsAhwILK+qdS0ELAfmtnmPqqpzq6qA04BDN9oGSZI0xQzsHIIk05KsBG6m+1I/v806vh0WOCHJtq1sV+D6nsXXtLL1la/pU96vHfOTrEiyYu3atQ96uyRJmgoGFgiq6p6qmg3MAPZN8jTgHcBTgGcCjwbe3qr3O/5fG1Derx0nV9Wcqpozffr0cW6FJElT08CvMqiqnwHfBOZW1Y3tsMCdwL8C+7Zqa4DdehabAdwwSvmMPuWSJGkMBnWVwfQkO7Tx7YAXAt9vx/5pVwQcClzRFlkCvKZdbbAfcGtV3QgsAw5IsmOSHYEDgGVt3m1J9mvreg1w5iC2TZKkqWBQVxnsApyaZBpdCDmjqr6c5BtJptN1+a8E/qLVXwocDKwG7gBeB1BV65IcB1zY6r27qta18b8EFgHb0V1d4BUGkiSN0UACQVVdBjyjT/nzR6hfwJEjzFsILOxTvgJ42oNrqSRJWybvVChJkgwEkiTJQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkSWJAgSDJQ5NckOTSJKuSHNvK90hyfpKrk/x7km1a+bZtenWbP7NnXe9o5VclObCnfG4rW53k6EFslyRJU8WgegjuBJ5fVU8HZgNzk+wHvA84oapmAbcAR7T6RwC3VNWTgBNaPZLsCRwG7AXMBT6WZFqSacBHgYOAPYHDW11JkjQGAwkE1bm9TW7dhgKeD3yulZ8KHNrGD2nTtPkvSJJWvriq7qyqa4HVwL5tWF1V11TVXcDiVleSJI3BwM4haL/kVwI3A8uBHwI/q6q7W5U1wK5tfFfgeoA2/1bgMb3lw5YZqVySJI3BwAJBVd1TVbOBGXS/6J/ar1p7zAjzxlv+AEnmJ1mRZMXatWtHb7gkSVuAgV9lUFU/A74J7AfskGSrNmsGcEMbXwPsBtDmbw+s6y0ftsxI5f2e/+SqmlNVc6ZPnz4RmyRJ0mZvUFcZTE+yQxvfDnghcCVwNvCyVm0ecGYbX9KmafO/UVXVyg9rVyHsAcwCLgAuBGa1qxa2oTvxcMnG3zJJkqaGrUavMiF2AU5tVwM8BDijqr6c5HvA4iTvAS4BTmn1TwE+lWQ1Xc/AYQBVtSrJGcD3gLuBI6vqHoAkRwHLgGnAwqpaNaBtkyRpszeQQFBVlwHP6FN+Dd35BMPLfwW8fIR1HQ8c36d8KbD0QTdWkqQtkHcqlCRJBgJJkmQgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkMaBAkGS3JGcnuTLJqiRvaOXHJPlxkpVtOLhnmXckWZ3kqiQH9pTPbWWrkxzdU75HkvOTXJ3k35NsM4htkyRpKhhUD8HdwJur6qnAfsCRSfZs806oqtltWArQ5h0G7AXMBT6WZFqSacBHgYOAPYHDe9bzvrauWcAtwBED2jZJkjZ7AwkEVXVjVV3cxm8DrgR2Xc8ihwCLq+rOqroWWA3s24bVVXVNVd0FLAYOSRLg+cDn2vKnAodunK2RJGnqGfg5BElmAs8Azm9FRyW5LMnCJDu2sl2B63sWW9PKRip/DPCzqrp7WLkkSRqDgQaCJI8APg+8sap+DpwEPBGYDdwIfGCoap/FawPK+7VhfpIVSVasXbt2nFsgSdLUNLBAkGRrujBwelV9AaCqbqqqe6rqXuATdIcEoPuFv1vP4jOAG9ZT/lNghyRbDSt/gKo6uarmVNWc6dOnT8zGSZK0mRvUVQYBTgGurKoP9pTv0lPtpcAVbXwJcFiSbZPsAcwCLgAuBGa1Kwq2oTvxcElVFXA28LK2/DzgzI25TZIkTSVbjV5lQjwHeDVweZKVreyddFcJzKbr3r8O+HOAqlqV5Azge3RXKBxZVfcAJDkKWAZMAxZW1aq2vrcDi5O8B7iELoBIkqQxGEggqKpv0/84/9L1LHM8cHyf8qX9lquqa7jvkIMkSRoH71QoSZIMBJIkyUAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJYhyBIMnLRyh/2cQ1R5IkTYbx9BCcMkL5yRPREEmSNHm2Gq1Ckie00Yck2QNIz+wnAL/aGA2TJEmDM2ogAFYDRRcEfjhs3k+AYya4TZIkacBGDQRV9RCAJP9VVc/b+E2SJEmDNuZzCAwDkiRNXWM5ZABAO3/geGA28IjeeVX1+AlulyRJGqAxBwLg03TnELwZuGPjNEeSJE2G8QSCvYDnVNW9G6sxkiRpcoznPgTnAM/YkCdJsluSs5NcmWRVkje08kcnWZ7k6va4YytPkhOTrE5yWZK9e9Y1r9W/Osm8nvJ9klzeljkxSR7YEkmS1M94egiuA5Yl+QLd5Yb/p6reNcqydwNvrqqLkzwSuCjJcuC1wFlV9d4kRwNHA28HDgJmteFZwEnAs5I8GlgAzKG7FPKiJEuq6pZWZz5wHrAUmAt8ZRzbJ0nSFms8PQQPB74EbA3sNmxYr6q6saoubuO3AVcCuwKHAKe2aqcCh7bxQ4DTqnMesEOSXYADgeVVta6FgOXA3DbvUVV1blUVcFrPuiRJ0ijG3ENQVa+biCdMMpPu0MP5wM5VdWNb/41JHtuq7Qpc37PYmla2vvI1fcr7Pf98up4EHv94L46QJAnGd9nhE0aaV1XXjHEdjwA+D7yxqn6+nsP8/WbUBpQ/sLDqZNr/X5gzZ07fOpIkbWnGcw5B7y2Mhwx9oU4bbeEkW9OFgdOr6gut+KYku7TegV2Am1v5Gu5/KGIGcEMr339Y+Tdb+Yw+9SVJ0hiM55DB/c43SPI4uhP8vjXasu2M/1OAK6vqgz2zlgDzgPe2xzN7yo9KspjupMJbW2hYBvzD0NUIwAHAO6pqXZLbkuxHdyjiNcCHx7ptkjRV3HrssZPdBE2g7RcsGNhzjaeH4H6q6idJ3gj8gO6mRevzHODVwOVJVrayd9IFgTOSHAH8CHh5m7cUOJiuV+IO4HXtOdclOQ64sNV7d1Wta+N/CSwCtqO7usArDCRJGqMNDgTNk4GHjVapqr5N/+P8AC/oU7+AI0dY10JgYZ/yFcDTRmuLJEl6oPGcVPgt7n+i3sPo7l747olulCRJGqzx9BB8ctj0L4BLq+rqCWyPJEmaBOM5qfDU0WtJkqTN0ZjvVJhk6yTHJrkmya/a47FJttmYDZQkSRvfeA4Z/D9gX+AvgP8Gdgf+HngU8KaJb5okSRqU8QSClwNPr6r/adNXJbkYuBQDgSRJm7Xx/HOjkS4b9N8MS5K0mRtPIPgs8KUkByZ5apK5wH+0ckmStBkbzyGDtwF/B3wU+DXgx8BngPdshHZJkqQBGrWHIMlzkryvqu6qqndV1ZOq6mFVNQvYFth74zdTkiRtTGM5ZPBO4JwR5p0N/O3ENUeSJE2GsQSC2cBXR5j3dWCfiWuOJEmaDGMJBI8CRrr50NbAIyeuOZIkaTKMJRB8HzhghHkHtPmSJGkzNparDE4APp5kGvAfVXVvkocAh9JdcfA3G7OBkiRp4xs1EFTVp5M8DjgV2DbJT4GdgF8BC6rqMxu5jZIkaSMb030IquqDST4J/BbwGOB/gHOr6ucbs3GSJGkwxvPvj38OLNuIbZEkSZNkPLculiRJU5SBQJIkGQgkSZKBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkMaBAkGRhkpuTXNFTdkySHydZ2YaDe+a9I8nqJFclObCnfG4rW53k6J7yPZKcn+TqJP+eZJtBbJckSVPFoHoIFgFz+5SfUFWz27AUIMmewGHAXm2ZjyWZlmQa8FHgIGBP4PBWF+B9bV2zgFuAIzbq1kiSNMUMJBBU1TnAujFWPwRYXFV3VtW1wGpg3zasrqprquouYDFwSJIAzwc+15Y/FTh0QjdAkqQpbrLPITgqyWXtkMKOrWxX4PqeOmta2UjljwF+VlV3DyuXJEljNJmB4CTgicBs4EbgA608ferWBpT3lWR+khVJVqxdu3Z8LZYkaYqatEBQVTdV1T1VdS/wCbpDAtD9wt+tp+oM4Ib1lP8U2CHJVsPKR3rek6tqTlXNmT59+sRsjCRJm7lJCwRJdumZfCkwdAXCEuCwJNsm2QOYBVwAXAjMalcUbEN34uGSqirgbOBlbfl5wJmD2AZJkqaKrUav8uAl+QywP7BTkjXAAmD/JLPpuvevA/4coKpWJTkD+B5wN3BkVd3T1nMUsAyYBiysqlXtKd4OLE7yHuAS4JRBbJckSVPFQAJBVR3ep3jEL+2qOh44vk/5UmBpn/JruO+QgyRJGqfJvspAkiRtAgwEkiTJQCBJkgwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkhhQIEiyMMnNSa7oKXt0kuVJrm6PO7byJDkxyeoklyXZu2eZea3+1Unm9ZTvk+TytsyJSTKI7ZIkaaoYVA/BImDusLKjgbOqahZwVpsGOAiY1Yb5wEnQBQhgAfAsYF9gwVCIaHXm9yw3/LkkSdJ6DCQQVNU5wLphxYcAp7bxU4FDe8pPq855wA5JdgEOBJZX1bqqugVYDsxt8x5VVedWVQGn9axLkiSNwWSeQ7BzVd0I0B4f28p3Ba7vqbemla2vfE2fckmSNEab4kmF/Y7/1waU9195Mj/JiiQr1q5du4FNlCRpapnMQHBT6+6nPd7cytcAu/XUmwHcMEr5jD7lfVXVyVU1p6rmTJ8+/UFvhCRJU8FkBoIlwNCVAvOAM3vKX9OuNtgPuLUdUlgGHJBkx3Yy4QHAsjbvtiT7tasLXtOzLkmSNAZbDeJJknwG2B/YKckauqsF3guckeQI4EfAy1v1pcDBwGrgDuB1AFW1LslxwIWt3rurauhExb+ku5JhO+ArbZAkSWM0kEBQVYePMOsFfeoWcOQI61kILOxTvgJ42oNpoyRJW7JN8aRCSZI0YAYCSZJkIJAkSQYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmShIFAkiRhIJAkSRgIJEkSBgJJkoSBQJIkYSCQJEkYCCRJEgYCSZKEgUCSJGEgkCRJGAgkSRIGAkmSBGw12Q2QdJ9bjz12spugCbT9ggWT3QRpzOwhkCRJBgJJkrQJBIIk1yW5PMnKJCta2aOTLE9ydXvcsZUnyYlJVie5LMnePeuZ1+pfnWTeZG2PJEmbo0kPBM3vVtXsqprTpo8GzqqqWcBZbRrgIGBWG+YDJ0EXIIAFwLOAfYEFQyFCkiSNblMJBMMdApzaxk8FDu0pP6065wE7JNkFOBBYXlXrquoWYDkwd9CNliRpc7UpBIICvpbkoiTzW9nOVXUjQHt8bCvfFbi+Z9k1rWyk8gdIMj/JiiQr1q5dO4GbIUnS5mtTuOzwOVV1Q5LHAsuTfH89ddOnrNZT/sDCqpOBkwHmzJnTt44kSVuaSe8hqKob2uPNwBfpzgG4qR0KoD3e3KqvAXbrWXwGcMN6yiVJ0hhMaiBI8vAkjxwaBw4ArgCWAENXCswDzmzjS4DXtKsN9gNubYcUlgEHJNmxnUx4QCuTJEljMNmHDHYGvphkqC2frqqvJrkQOCPJEcCPgJe3+kuBg4HVwB3A6wCqal2S44ALW713V9W6wW2GJEmbt0kNBFV1DfD0PuX/A7ygT3kBR46wroXAwoluoyRJW4JJP4dAkiRNPgOBJEkyEEiSJAOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJAwEkiQJA4EkScJAIEmSMBBIkiQMBJIkCQOBJEnCQCBJkjAQSJIkDASSJAkDgSRJwkAgSZIwEEiSJKZYIEgyN8lVSVYnOXqy2yNJ0uZiygSCJNOAjwIHAXsChyfZc3JbJUnS5mHKBAJgX2B1VV1TVXcBi4FDJrlNkiRtFqZSINgVuL5nek0rkyRJo0hVTXYbJkSSlwMHVtWftulXA/tW1V8PqzcfmN8mnwxcNdCGTg07AT+d7EZos+Y+pAfLfWjD7F5V0/vN2GrQLdmI1gC79UzPAG4YXqmqTgZOHlSjpqIkK6pqzmS3Q5sv9yE9WO5DE28qHTK4EJiVZI8k2wCHAUsmuU2SJG0WpkwPQVXdneQoYBkwDVhYVasmuVmSJG0WpkwgAKiqpcDSyW7HFsBDLnqw3If0YLkPTbApc1KhJEnacFPpHAJJkrSBDASbsCSV5AM9029Jcswoyxw60h0akxyT5MdJVib5XpLDJ7Ct1yXZaRz1XzJ0e+nhbU7y2iS/Ns7nn5nkivEssyVIckKSN/ZML0vyyZ7pDyT5myS/luRzrWx2koN76hyT5C0jrP+742zPcUkua/vg14be53RObLcdvyzJ3iMsf09bdlWSS1vb/RybBEle2j6jnjKs/P3t/Xn/+j6PxvE8+7b3fGV7z1/aM2/U29UnWZTk2rbsD5KclsR71PThH9Km7U7gD8bzRQscSnfr5pGcUFWz6e7i+PEkWz+YBm6oqlpSVe9tk8Pb/FpgXIFAI/ou8GyA9sW5E7BXz/xnA9+pqhuq6mWtbDZwMGNQVc8eZ3veX1W/2fbBLwPvauUHAbPaMB84aYTlf1lVs6tqL+D3WjsXjLMNmhiHA9+mu6Kr158De1fVWxn98+gBkgw/t+0KYE7bZ+bSfW5tNc7b1b+1qp5Od++ZS4Cz29Vo6mEg2LTdTXfizJuGz0iye5Kz2q+ps5I8PsmzgZcA729p+okjrbiqrgbuAHZs63tikq8muSjJt4ZSf5IXJzk/ySVJvp5k51b+mPYL75IkHwfSymcm+X6STya5IsnpSV6Y5DtJrk6yb6v32iQf6dPmtwNzgNPb9HZJ9knyX61ty5Ls0taxT0v95wJHTnAbHW0AAAbTSURBVMxLPuV8hxYI6ILAFcBtSXZMsi3wVOCSoR6W9iH5buAV7fV/RVt2zyTfTHJNktcPrTzJ7e1x/zb/c+39Pz1Jhjemqn7eM/lwYOgkpkOA06pzHrDD0Ps8kqq6mS48HNV6GGa2fffiNgwFoU8l+b/bmLe2vSTJXkkuaNt5WZJZY3tJleQRwHOAI+gJBEmW0L2v5ydZwLDPo/V8zixK8sEkZwPv632uqrqjqu5ukw/lvn1m3Lerb/vXCcBP6IIESU5KsqL1ahzbyl6Q5Is92/V7Sb6QZFpr6xVJLk/ygM/mzVpVOWyiA3A78CjgOmB74C3AMW3el4B5bfxPgP9o44uAl42wvmOAt7TxvYFv9cw7C5jVxp8FfKON78h9J5/+KfCBNn4i8K42/iK6P9KdgJl0QeY36ALnRcBCusBwSE87Xwt8pF+bgW/S/SIA2JruV+70Nv0KuktKAS4DntfG3w9cMdnv2aY4tP3n8XS/3P4COI7ul/VzgHNanZlDr1/ve9Oz33wX2La9x/8DbD20j7bH/YFb6W4I9hDgXOC5I7TneLrbjF/R875+ubd+2x/n9Pub6FN2C7Az8DDgoa1sFrCijT+vZ7/bHriW7gqrDwOvbOXbANtN9nu1uQzAq4BT2vh36XoEHvAe9fnbHulzZlHbB6aN8HzPAlbRfSa+tJW9DPhkT51X9+63I7WhlX0IeHsbf3R7nEb32fObdJ9X3+/ZPz8NvBjYB1jes54dJvu9mMjBHoJNXHW/qE4DXj9s1m/R7aQAnwKeO8ZVvinJVcD5dB/0Q2n/2cBnk6wEPg4M/TqbASxLcjnwVu7rbv4d4N9aG/+T7kN5yLVVdXlV3Uv3R3xWdX89l9N98YzHk4GnActb2/4OmJFke7o/xv9q9T41zvVuSYZ6CZ5N90V9bs/0WM8B+M+qurOqfgrcTPcFPNwFVbWmve8rGeG9rqq/rardgNOBo1rxA3oTuO+X4GiGlt0a+ETbVz9L66pu+8iTkjyWrpv789X94jwXeGfrldq9qn45xudT9zoubuOL2/R6jfI5A/DZqrqn37JVdX51h4meCbwjyUOZmH0G4I+SXEx3KGEvYM/2efUp4FVJdqD7vP0KcA3whCQfTjIX+DlTyJS6D8EU9iHgYuBf11NnrH8IJ1TVPyX5A+C0dljhIcDPqjtGN9yHgQ9W1ZIk+9NCxCjPeWfP+L090/cy/n0uwKqq+q37FXZ/pF4zOzZD5xH8Bt2v8uuBN9N9mC0c4zp639N76P8+jqVOr08D/0l3DsCYbj0+XJIntOe6ua3nJuDpdPv0r3qqfgp4JV339p8AVNWnk5xP18O1LMmfVtU3RnvOLV2SxwDPB56WpOh+WVeSt7Uv0pGs73MG4BejPXdVXZnkF3Q/EjZon2meAZyVZA+6ntdnVtUtSRbRHZaA7vP2S3T70WdbiLwlydOBA+kOU/4RbX+aCuwh2AxU1TrgDLrjdUO+y33H7l5Jd3IPwG3AI8ewzi8AK+gOO/wcuDbdP4gaOuP76a3q9sCP2/i8nlWc056XJAfRzkXYQMPb3Dt9FTA9yW+159o6yV5V9TPg1iRDPSOvfBDPP9V9B/h9YF1V3dP2p6FfPef2qT+mfWhDDDtO/xK6blnobjP+mrbv7QfcWlU3jrKu6cC/0HUTF92+emProXg13RfVkEXAGwGq3cG0hYlrqurE9vy/+SA3b0vxMrrzPXavqpmtt+da+vdS/t++NMrnzIjS3Y5+qza+O12v4XVswO3q23O+nq5n4qt0h2R/QfdZsjPtvILW3hvoAsbf0e0/pDvB+yFV9Xng7+kOvU4ZBoLNxwfojt8OeT3wuiSX0X34vaGVLwbemu5kvxFPKmzeDQxdtvVK4Igkl9J18w+dnHMMXRfft7j/fxY7Fvid1tV2APCjDd6yB7Z5EfAvrVtxGt0H0Pta21Zy30lyrwM+mu6kQrt7R3Y53b5z3rCyW9shgOHOpjuJsPekwony3nZC1mV0+83QfruUrjt2NfAJ4K9GWH671q5VwNeBr9HtiwAfA+YlOQ/4dXp+cVbVTcCV3L+X7RXAFW0/ewrdoTmN7nDgi8PKPg/8cZ+6w/+2R/qcWZ/nApe29+mLwF9V1U/bL/ah29VfCZxRI9+u/v3tOX9Ad9jhd6vqrqq6lO5QwSq63rLvDFvudOD6qvpem94V+GZryyLgHWNo/2bDOxVKmvKSPIwuBO1dVbdOdnu0eUjyEeCSqjplstsyCPYQSJrSkryQ7tDEhw0DGqskF9EdRvq3yW7LoNhDIEmS7CGQJEkGAkmShIFAkiRhIJAkSRgIJEkSBgJJkgT8f3jFNzlqG2jaAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(8,8))\n",
    "plt.bar(x, y, color ='lightcoral')\n",
    "plt.title('Readmission Rate', fontsize=14)\n",
    "plt.ylabel('Count', fontsize=12)\n",
    "# plt.savefig('Readmit_rate.png')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, we looked at age. From the literature and the below graph, we see that age has a non-linear effect on readmission. Patients 0-30 years old are rarely readmitted. Patients 30-60 are more likely to be readmitted. Then as patients reach 60-100, this value decreases. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "age_prct = pd.DataFrame(df.groupby(by = 'age')['readmitted'].value_counts())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7fe6f24da250>"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "age_prct.unstack().plot(kind='line', color=['maroon', 'lightcoral', 'firebrick'], figsize=(6,4))\n",
    "plt.xlabel('Age', fontsize=12)\n",
    "plt.ylabel('Count',fontsize=12)\n",
    "plt.title('Readmission Rate by Age', fontsize=14)\n",
    "plt.legend(['Not Readmitted', 'Within 30 Days', 'After 30 Days'])\n",
    "# plt.savefig('Readmit_vs_age.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, we looked at how the patient's race affects readmission. From the American Diabetes Assoc, Black and Hispanic populations have a greater incidence of diabetes than Asian and Caucasian populations. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "race_prct = pd.DataFrame(df.groupby(by = 'race')['readmitted'].value_counts(normalize = True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7fe6f8dde290>"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "race_prct.unstack().plot(kind='bar', color=['maroon', 'lightcoral', 'firebrick'], stacked=True)\n",
    "plt.xlabel('Race', fontsize=12)\n",
    "plt.xticks(rotation=0)\n",
    "plt.ylabel('Probability',fontsize=12)\n",
    "plt.title('Readmission Rate by Race', fontsize=14)\n",
    "plt.legend(['Not Readmitted', 'Within 30 Days', 'After 30 Days'],loc=(1.01,-.01))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dummy Classifier before Feature Engineering"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.drop('readmitted', axis = 1)\n",
    "y = df.readmitted"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Dummy Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dummy Model Accuracy:  0.46342862606367724\n"
     ]
    }
   ],
   "source": [
    "dummy_clf = DummyClassifier()\n",
    "dummy_clf.fit(X, y)\n",
    "dummy_clf.predict(X)\n",
    "print(\"Dummy Model Accuracy: \", dummy_clf.score(X, y))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Feature Engineering"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Was the patient given a A1C test at all?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['A1C_test'] = np.where(df.A1Cresult == 'None', 0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's visualize this distribution of HbA1c tests."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "a1c_prct = df.A1C_test.value_counts(normalize=True) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Text(0.5, 1.0, 'HbA1c Test Coverage')"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "a1c_prct.plot(kind = 'bar', color = 'lightcoral', figsize=(6,6))\n",
    "plt.xlabel('A1C Test', fontsize=12)\n",
    "plt.xticks([0,1,],labels= ['Not Tested', 'Tested'])\n",
    "plt.xticks(rotation=0)\n",
    "plt.ylabel('Frequency',fontsize=12)\n",
    "plt.title('HbA1c Test Coverage', fontsize=14)\n",
    "#plt.savefig('Images/HbA1c_test.png')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Was the patient's meds changed during the hospitalization?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.change = np.where(df.change == 'No', 0, 1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Was the patient tested and meds changed?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['A1C_test_and_changed'] = np.where((df.change == 1) & (df.A1C_test == 1), 1, 0)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "From domain knowledge, patients are readmitted at different rates based on their age brackets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "conditions = [\n",
    "    (df.age ==  '[0-10)') | (df.age == '[10-20)') | (df.age == '[20-30)'),\n",
    "    (df.age == '[30-40)') | (df.age == '[40-50)') | (df.age == '[50-60)'),\n",
    "    (df.age == '[60-70)') | (df.age == '[70-80)') | (df.age == '[80-90)') | (df.age == '[90-100')]\n",
    "\n",
    "choices = [\n",
    "    '[0-30)',\n",
    "    '[30-60]',\n",
    "    '[60-100)']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['binned_age'] = np.select(conditions, choices, default=np.nan)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = df[df.binned_age != 'nan']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['age'], axis=1, inplace=True) # Dropping for correlation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Is diabetes one of the three primary diagnoses for the hospitalization?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "df['diabetes_as_diag_1'] = np.where((df.diag_1 >= 250) & (df.diag_1 <251), 1, 0)\n",
    "df['diabetes_as_diag_2'] = np.where((df.diag_2 >= 250) & (df.diag_2 <251), 1, 0)\n",
    "df['diabetes_as_diag_3'] = np.where((df.diag_3 >= 250) & (df.diag_3 <251), 1, 0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.drop(['diag_1', 'diag_2', 'diag_3'], axis=1, inplace=True) # Dropping for correlation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To pare down some features, we removed drugs that had fewer than 5000 (1/12 of the total) patients using them. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "meds_to_remove = ['repaglinide', 'nateglinide', 'chlorpropamide', 'glimepiride', 'acetohexamide', 'tolbutamide', \n",
    "            'pioglitazone', 'rosiglitazone', 'acarbose', 'miglitol', 'troglitazone', 'tolazamide', 'examide', \n",
    "            'citoglipton', 'glyburide-metformin', 'glipizide-metformin', 'glimepiride-pioglitazone', \n",
    "            'metformin-rosiglitazone', 'metformin-pioglitazone']\n",
    "df.drop(meds_to_remove, axis=1, inplace=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 28)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "How many patients had diabetes as one of the top 3 diagnoses for inpatient visit?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Number of patients with diabetes as top3 diagnosis:  24281\n"
     ]
    }
   ],
   "source": [
    "print(\"Number of patients with diabetes as top3 diagnosis: \", \n",
    "      11495 + 8227 + 4559) # Yes, for diag3, diag2, diag1, respectively"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percent of inpatient visits with diabetes listed in top 3 diagnoses:  39.788611224907825\n"
     ]
    }
   ],
   "source": [
    "print(\"Percent of inpatient visits with diabetes listed in top 3 diagnoses: \", 24281 / len(df) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percent of patients who received HbA1c test:  18.55960671855797\n"
     ]
    }
   ],
   "source": [
    "print(\"Percent of patients who received HbA1c test: \", len(df[df.A1C_test == 1]) / len(df) * 100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Percent of patients who change meds after HbA1c tesT:  0.5580964153275649\n"
     ]
    }
   ],
   "source": [
    "print(\"Percent of patients who change meds after HbA1c tesT: \", \n",
    "      len(df[df.A1C_test_and_changed == 1]) / len(df[df.A1C_test == 1]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Modeling"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We printed 4 major metrics to get a better idea of how our model was performing, but we focused on F1 score as our main metric. We wanted a more well-rounded model, and did not see how false positives or false negatives would be relatively worse in either case.\n",
    "\n",
    "Because this is multiclass, average had to set to weighted."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "def print_metrics(labels, preds):\n",
    "    print(\"Precision Score: {}\".format(precision_score(labels, preds, average='weighted')))\n",
    "    print(\"Recall Score: {}\".format(recall_score(labels, preds, average='weighted')))\n",
    "    print(\"Accuracy Score: {}\".format(accuracy_score(labels, preds)))\n",
    "    print(\"F1 Score: {}\".format(f1_score(labels, preds, average='weighted')))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression: No Poly/Interaction Features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This was the best performing model and was used for further analysis/presentation. Iterative steps are included below these results and include many models of both logistic regression and decision trees."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Separate into features and target feature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.drop('readmitted', axis = 1)\n",
    "y = df.readmitted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "continuous_variables = ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', \n",
    "                       'number_outpatient', 'number_emergency', 'number_inpatient', 'number_diagnoses']\n",
    "categorical_variables = [x for x in X.columns if x not in continuous_variables]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cont = X[continuous_variables]\n",
    "df_cate = X[categorical_variables]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[60-100)    40635\n",
       "[30-60]     19195\n",
       "[0-30)       1195\n",
       "Name: binned_age, dtype: int64"
      ]
     },
     "execution_count": 75,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_cate.binned_age.value_counts()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Get dummy variables for categorical variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "dummy_df_cate = pd.get_dummies(df_cate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 45)"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "dummy_df_cate.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Concatenating our cont and cate features back together."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_all = pd.concat([df_cont, dummy_df_cate], axis = 1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 53)"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_all.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Test Train Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set - Features:  (45768, 53) Target:  (45768,)\n",
      "Training set - Features:  (15257, 53) Target:  (15257,)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df_all, y, random_state=30)\n",
    "print(\"Training set - Features: \", X_train.shape, \"Target: \", y_train.shape)\n",
    "print(\"Training set - Features: \", X_test.shape, \"Target: \",y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Running Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "logreg_nopoly = LogisticRegressionCV(random_state=30, max_iter=1000)\n",
    "model_log_nopoly = logreg_nopoly.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model_log_nopoly.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.8988571301977448\n",
      "Recall Score: 0.6024120076030675\n",
      "Accuracy Score: 0.6024120076030675\n",
      "F1 Score: 0.7061286748857043\n"
     ]
    }
   ],
   "source": [
    "print_metrics(y_pred, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [],
   "source": [
    "pickle.dump(logreg_nopoly, open('diabetes_model.pkl', 'wb'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "results = pd.DataFrame(model_log_nopoly.coef_, columns=X_train.columns).T"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "This dataframe has the coefficients for the three classes (0,1,2). Selecting just to see the highest values."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "results = results[(results > .08) | (results < -.08)]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression: Poly/Interaction Features"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Does including polynomial and interaction features improve the model?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "X = df.drop('readmitted', axis = 1)\n",
    "y = df.readmitted"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generating Polynomical/Interaction Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "continuous_variables = ['time_in_hospital', 'num_lab_procedures', 'num_procedures', 'num_medications', \n",
    "                       'number_outpatient', 'number_emergency', 'number_inpatient', 'number_diagnoses']\n",
    "categorical_variables = [x for x in X.columns if x not in continuous_variables]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_cont = X[continuous_variables]\n",
    "df_cate = X[categorical_variables]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "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>race</th>\n",
       "      <th>gender</th>\n",
       "      <th>admission_type_id</th>\n",
       "      <th>discharge_disposition_id</th>\n",
       "      <th>admission_source_id</th>\n",
       "      <th>max_glu_serum</th>\n",
       "      <th>A1Cresult</th>\n",
       "      <th>metformin</th>\n",
       "      <th>glipizide</th>\n",
       "      <th>glyburide</th>\n",
       "      <th>insulin</th>\n",
       "      <th>change</th>\n",
       "      <th>diabetesMed</th>\n",
       "      <th>A1C_test</th>\n",
       "      <th>A1C_test_and_changed</th>\n",
       "      <th>binned_age</th>\n",
       "      <th>diabetes_as_diag_1</th>\n",
       "      <th>diabetes_as_diag_2</th>\n",
       "      <th>diabetes_as_diag_3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Female</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Up</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[0-30)</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Male</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Up</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[30-60]</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Male</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[30-60]</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Male</td>\n",
       "      <td>2</td>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[30-60]</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Male</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[60-100)</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101754</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Female</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>&gt;7</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Up</td>\n",
       "      <td>Steady</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>[60-100)</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101755</th>\n",
       "      <td>Other</td>\n",
       "      <td>Female</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>&gt;8</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>No</td>\n",
       "      <td>Up</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>[30-60]</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101756</th>\n",
       "      <td>Other</td>\n",
       "      <td>Female</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Steady</td>\n",
       "      <td>0</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[60-100)</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101758</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Female</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>Up</td>\n",
       "      <td>1</td>\n",
       "      <td>Yes</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[60-100)</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>101765</th>\n",
       "      <td>Caucasian</td>\n",
       "      <td>Male</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>7</td>\n",
       "      <td>None</td>\n",
       "      <td>None</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>No</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>[60-100)</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>61025 rows × 19 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "             race  gender  admission_type_id  discharge_disposition_id  \\\n",
       "1       Caucasian  Female                  1                         1   \n",
       "3       Caucasian    Male                  1                         1   \n",
       "4       Caucasian    Male                  1                         1   \n",
       "5       Caucasian    Male                  2                         1   \n",
       "7       Caucasian    Male                  1                         1   \n",
       "...           ...     ...                ...                       ...   \n",
       "101754  Caucasian  Female                  1                         1   \n",
       "101755      Other  Female                  1                         1   \n",
       "101756      Other  Female                  1                         1   \n",
       "101758  Caucasian  Female                  1                         1   \n",
       "101765  Caucasian    Male                  1                         1   \n",
       "\n",
       "        admission_source_id max_glu_serum A1Cresult metformin glipizide  \\\n",
       "1                         7          None      None        No        No   \n",
       "3                         7          None      None        No        No   \n",
       "4                         7          None      None        No    Steady   \n",
       "5                         2          None      None        No        No   \n",
       "7                         7          None      None        No        No   \n",
       "...                     ...           ...       ...       ...       ...   \n",
       "101754                    7          None        >7        No        No   \n",
       "101755                    7          None        >8        No    Steady   \n",
       "101756                    7          None      None        No        No   \n",
       "101758                    7          None      None        No        No   \n",
       "101765                    7          None      None        No        No   \n",
       "\n",
       "       glyburide insulin  change diabetesMed  A1C_test  A1C_test_and_changed  \\\n",
       "1             No      Up       1         Yes         0                     0   \n",
       "3             No      Up       1         Yes         0                     0   \n",
       "4             No  Steady       1         Yes         0                     0   \n",
       "5             No  Steady       0         Yes         0                     0   \n",
       "7         Steady      No       0         Yes         0                     0   \n",
       "...          ...     ...     ...         ...       ...                   ...   \n",
       "101754        Up  Steady       1         Yes         1                     1   \n",
       "101755        No      Up       1         Yes         1                     1   \n",
       "101756        No  Steady       0         Yes         0                     0   \n",
       "101758        No      Up       1         Yes         0                     0   \n",
       "101765        No      No       0          No         0                     0   \n",
       "\n",
       "       binned_age  diabetes_as_diag_1  diabetes_as_diag_2  diabetes_as_diag_3  \n",
       "1          [0-30)                   0                   1                   0  \n",
       "3         [30-60]                   0                   1                   0  \n",
       "4         [30-60]                   0                   0                   1  \n",
       "5         [30-60]                   0                   0                   1  \n",
       "7        [60-100)                   0                   0                   1  \n",
       "...           ...                 ...                 ...                 ...  \n",
       "101754   [60-100)                   0                   0                   1  \n",
       "101755    [30-60]                   0                   0                   0  \n",
       "101756   [60-100)                   0                   0                   0  \n",
       "101758   [60-100)                   0                   0                   0  \n",
       "101765   [60-100)                   0                   0                   0  \n",
       "\n",
       "[61025 rows x 19 columns]"
      ]
     },
     "execution_count": 77,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_cate"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generating polynomial and interaction features for continuous variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "poly_cont = PolynomialFeatures(2)\n",
    "poly_df_cont_data = poly_cont.fit_transform(df_cont)\n",
    "poly_df_cont_cols = poly_cont.get_feature_names(df_cont.columns)\n",
    "poly_df_cont = pd.DataFrame(poly_df_cont_data, columns=poly_df_cont_cols) # 45 features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 45)"
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "poly_df_cont.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Generating interaction features for categorical variables."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "dummy_df_cate = pd.get_dummies(df_cate)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "poly_cate = PolynomialFeatures(interaction_only=True)\n",
    "poly_df_cate_data = poly_cate.fit_transform(dummy_df_cate)\n",
    "poly_df_cate_cols = poly_cate.get_feature_names(dummy_df_cate.columns)\n",
    "poly_df_cate = pd.DataFrame(poly_df_cate_data, columns=poly_df_cate_cols) # 1,036 features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 1036)"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "poly_df_cate.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Concatenating our cont and cate features back together."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_all = pd.concat([poly_df_cont, poly_df_cate], axis = 1) # 1,081 features total"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(61025, 1081)"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df_all.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train Test Split"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training set - Features:  (45768, 1081) Target:  (45768,)\n",
      "Training set - Features:  (15257, 1081) Target:  (15257,)\n"
     ]
    }
   ],
   "source": [
    "X_train, X_test, y_train, y_test = train_test_split(df_all, y, random_state=30)\n",
    "print(\"Training set - Features: \", X_train.shape, \"Target: \", y_train.shape)\n",
    "print(\"Training set - Features: \", X_test.shape, \"Target: \",y_test.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Scaling our data: fit to the training set then apply to both train and test sets."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "scaler = StandardScaler()\n",
    "X_train = pd.DataFrame(scaler.fit_transform(X_train), columns = X_train.columns)\n",
    "X_test = pd.DataFrame(scaler.transform(X_test), columns = X_train.columns)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Removing Correlated Features\n",
    "\n",
    "Want to remove correlated features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "corr_matrix = X_train.corr().abs()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [],
   "source": [
    "upper = corr_matrix.where(np.triu(np.ones(corr_matrix.shape), k=1).astype(np.bool))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Num of Removed Features:  428\n"
     ]
    }
   ],
   "source": [
    "correlated = [column for column in upper.columns if any(upper[column] > 0.75)]\n",
    "print(\"Num of Removed Features: \", len(correlated))\n",
    "X_train.drop(columns=correlated, inplace=True)\n",
    "X_test.drop(columns=correlated, inplace=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Running the Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "logreg = LogisticRegressionCV(random_state=30, max_iter=1000)\n",
    "model_log = logreg.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_pred = model_log.predict(X_test)\n",
    "y_pred_train = model_log.predict(X_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Test metrics:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.878476112823396\n",
      "Recall Score: 0.6000524349478927\n",
      "Accuracy Score: 0.6000524349478927\n",
      "F1 Score: 0.6974500889102002\n"
     ]
    }
   ],
   "source": [
    "print_metrics(y_pred, y_test)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Training metrics to ensure not overfitting:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.8855551820076031\n",
      "Recall Score: 0.6088751966439434\n",
      "Accuracy Score: 0.6088751966439434\n",
      "F1 Score: 0.7051051372857531\n"
     ]
    }
   ],
   "source": [
    "print_metrics(y_pred_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Just grabbing the top 10 positive and negative features because there are 1000 total features."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "info = list(tuple(zip(X_train.columns, model_log.coef_[0])))\n",
    "info.sort(key = lambda x: x[1]) \n",
    "top_10_pos = info[-10:]\n",
    "top_10_neg = info[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('number_inpatient', -0.1638442434818923),\n",
       " ('number_diagnoses', -0.10233336085419192),\n",
       " ('admission_source_id max_glu_serum_None', -0.08254081093691162),\n",
       " ('number_emergency', -0.07650544064795876),\n",
       " ('number_outpatient', -0.059371989527597056),\n",
       " ('time_in_hospital', -0.05603253101460419),\n",
       " ('discharge_disposition_id binned_age_[30-60]', -0.032153581466377666),\n",
       " ('num_procedures number_emergency', -0.030348182515057477),\n",
       " ('discharge_disposition_id diabetes_as_diag_1', -0.03021334118394899),\n",
       " ('discharge_disposition_id metformin_Steady', -0.02695735377818883)]"
      ]
     },
     "execution_count": 67,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_10_neg"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('admission_type_id max_glu_serum_None', 0.01828047882318645),\n",
       " ('diabetes_as_diag_3 metformin_Steady', 0.019380333101187537),\n",
       " ('admission_source_id change', 0.019397825725268644),\n",
       " ('diabetes_as_diag_3 metformin_Up', 0.020870950494928382),\n",
       " ('glyburide_Steady binned_age_[30-60]', 0.022283443671238974),\n",
       " ('A1C_test diabetes_as_diag_1', 0.027364724279092572),\n",
       " ('number_outpatient^2', 0.028064394774789496),\n",
       " ('discharge_disposition_id admission_source_id', 0.029890207791643874),\n",
       " ('num_procedures', 0.03529798221610847),\n",
       " ('number_emergency^2', 0.04668965402567975)]"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "top_10_pos"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GridSearch CV\n",
    "\n",
    "Tested Cs (0.01, 0.1, 1, 10, 20), cv (1, 3, 5, 7), penalty (L1, L2), and max_iter (100, 500, 1000), but best estimator performed worse than default by 2%."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [],
   "source": [
    "# logreg_grid = LogisticRegressionCV(random_state=30, max_iter=1000, n_jobs=-1)\n",
    "\n",
    "# param_grid ={'Cs': [1, 10, 20], \n",
    "#             'cv': [1, 3, 5, 7]}\n",
    "\n",
    "# #create the instance of GridSearchCV using the F1 metric for our scoring. \n",
    "# scorer = make_scorer(f1_score, average='weighted')\n",
    "# logreg_grid=GridSearchCV(logreg_grid, param_grid, scoring=scorer, verbose=2, n_jobs=-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [],
   "source": [
    "# logreg_grid.fit(X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [],
   "source": [
    "# y_pred = logreg_grid.best_estimator_.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [],
   "source": [
    "# print(\"Best Parameter Combination Found During Grid Search:\")\n",
    "# logreg_grid.best_params_"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [],
   "source": [
    "# print_metrics(y_test, y_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Logistic Regression with SMOTE\n",
    "\n",
    "There is some class imbalance in the set, but SMOTE decreased F1 score by 1%. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [],
   "source": [
    "# from imblearn.over_sampling import SMOTE\n",
    "# sm = SMOTE(random_state = 2)\n",
    "# X_train_smote, y_train_smote = sm.fit_sample(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [],
   "source": [
    "# X_train_smote.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 76,
   "metadata": {},
   "outputs": [],
   "source": [
    "# logreg = LogisticRegressionCV(random_state=30, max_iter=1000)\n",
    "# model_log_smote = logreg.fit(X_train_smote, y_train_smote)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [],
   "source": [
    "# y_pred = model_log_smote.predict(X_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "# print_metrics(y_pred, y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Recursive Feature Elimination\n",
    "\n",
    "RFE ran with over 600 features, but did not improve F1 score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "metadata": {},
   "outputs": [],
   "source": [
    "# rfe_log = LogisticRegressionCV(random_state=30, max_iter=1000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Instantiate a recursive eliminator object\n",
    "# selector = RFECV(verbose=1, estimator=rfe_log, step=10, scoring='neg_mean_squared_error')\n",
    "\n",
    "# # Fit recursive feature eliminator \n",
    "# selector.fit(X_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "metadata": {},
   "outputs": [],
   "source": [
    "# selected_rfe = X_train.columns[selector.support_]\n",
    "# removed_rfe = X_train.columns[~selector.support_]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "metadata": {},
   "outputs": [],
   "source": [
    "# len(list(removed_rfe)) # We eliminated 30 features from our model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 83,
   "metadata": {},
   "outputs": [],
   "source": [
    "# list(selected_rfe) # Here are the features we are keeping"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Decision Tree"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Baseline Decision Tree"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.49161513184402705\n",
      "Recall Score: 0.48515435537785934\n",
      "Accuracy Score: 0.48515435537785934\n",
      "F1 Score: 0.4882869868868028\n"
     ]
    }
   ],
   "source": [
    "clf = DecisionTreeClassifier()\n",
    "model_dt = clf.fit(X_train,y_train)\n",
    "\n",
    "y_pred_dt_test = model_dt.predict(X_test)\n",
    "\n",
    "print_metrics(y_test,y_pred_dt_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree with Class Weight Balanced"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.49547026946827905\n",
      "Recall Score: 0.48816936488169366\n",
      "Accuracy Score: 0.48816936488169366\n",
      "F1 Score: 0.4917001424323169\n"
     ]
    }
   ],
   "source": [
    "clf_balanced = DecisionTreeClassifier(class_weight = 'balanced')\n",
    "model_dt1 = clf_balanced.fit(X_train,y_train)\n",
    "\n",
    "y_pred_dt1_test = model_dt1.predict(X_test)\n",
    "\n",
    "print_metrics(y_test,y_pred_dt1_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree with SMOTE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 86,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Precision Score: 0.49760525476775597\n",
      "Recall Score: 0.4900045880579406\n",
      "Accuracy Score: 0.4900045880579406\n",
      "F1 Score: 0.49367378878518714\n"
     ]
    }
   ],
   "source": [
    "sm = SMOTE(sampling_strategy='minority', random_state=23)\n",
    "X_train_smote, y_train_smote = sm.fit_sample(X_train, y_train)\n",
    "smote_dtc = DecisionTreeClassifier()\n",
    "smote_dtc.fit(X_train_smote, y_train_smote)\n",
    "smote_pred = smote_dtc.predict(X_test)\n",
    "print_metrics(y_test, smote_pred)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Decision Tree with Tomek Links\n",
    "\n",
    "Commented out because took a long time to process, improved F1 score by 1%, but not better than logistic regression model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {},
   "outputs": [],
   "source": [
    "# tl = TomekLinks()\n",
    "# X_res, y_res = tl.fit_resample(X_train, y_train)\n",
    "\n",
    "# tomek_dtc = DecisionTreeClassifier()\n",
    "# tomek_dtc.fit(X_res, y_res)\n",
    "# tomek_pred = tomek_dtc.predict(X_test)\n",
    "# print_metrics(y_test, tomek_pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 88,
   "metadata": {},
   "outputs": [],
   "source": [
    "# F1: 0.50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### GridSearchCV Decision Tree\n",
    "\n",
    "Commented out because took a long time to process and did not improve F1 score."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 89,
   "metadata": {},
   "outputs": [],
   "source": [
    "# param_grid = {\n",
    "#     'max_depth': range(1,31,1),\n",
    "#     'criterion': ['gini','entropy'],\n",
    "#     'splitter' : ['best','random']}\n",
    "# scoring = {'f1':'f1_weighted', 'Accuracy': 'accuracy'}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 90,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dtc_grid = DecisionTreeClassifier(class_weight='balanced')\n",
    "# grid_tree = GridSearchCV(dtc_grid, param_grid, cv=5, scoring=scoring, refit= 'f1', verbose=1)\n",
    "# grid_tree.fit(X_train,y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 91,
   "metadata": {},
   "outputs": [],
   "source": [
    "# print(grid_tree.best_score_) # 0.4829 F1 Score\n",
    "# print(grid_tree.best_params_) # {'criterion': 'gini', 'max_depth': 28, 'splitter': 'random'}"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Insights"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We see from our best logistic regression model that many factors associated with diabetes and its treatment impact the rate of readmission. Unsurprisingly, if diabetes was listed as a main cause for the visit, then a patient was more likely to be readmitted within and after 30 days. These coefficients are more difficult to comprehend than with linear regression. A one unit increase in the feature corresponds to a change in the log odds of the class. What we can directly compare is the direction and magnitude of the relative changes.\n",
    "\n",
    "For not being readmitted, the most influential factors were if you were not on diabetic medication and if you were young (0-30 years old). If you had diabetic medication, had visited inpatient in the last year, or were between 60-100 years old, you were less likely to not be readmitted (sorry for the double-negative).\n",
    "\n",
    "For being readmitted within 30 days, previous inpatient visits, being 60-100 years old, and having diabetes as a main diagnosis increased the log odds. Interestingly, if you did not have a blood sugar test you were less likely to be readmitted within 30 days. However, you were more likley to be readmitted after 30 days. Suggesting that patients should have their blood sugar tested every visit because complications are not caught within 30 days but still arise. "
   ]
  }
 ],
 "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": "3.7.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}