mainPipeline.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "available-albany",
   "metadata": {},
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "import ipywidgets as widgets\n",
    "import sys\n",
    "from pathlib import Path\n",
    "import os\n",
    "import importlib\n",
    "\n",
    "\n",
    "module_path='preprocessing/day_intervals_preproc'\n",
    "if module_path not in sys.path:\n",
    "    sys.path.append(module_path)\n",
    "\n",
    "module_path='utils'\n",
    "if module_path not in sys.path:\n",
    "    sys.path.append(module_path)\n",
    "    \n",
    "module_path='preprocessing/hosp_module_preproc'\n",
    "if module_path not in sys.path:\n",
    "    sys.path.append(module_path)\n",
    "    \n",
    "module_path='model'\n",
    "if module_path not in sys.path:\n",
    "    sys.path.append(module_path)\n",
    "#print(sys.path)\n",
    "root_dir = os.path.dirname(os.path.abspath('UserInterface.ipynb'))\n",
    "import day_intervals_cohort\n",
    "from day_intervals_cohort import *\n",
    "\n",
    "import day_intervals_cohort_v2\n",
    "from day_intervals_cohort_v2 import *\n",
    "\n",
    "import data_generation_icu\n",
    "\n",
    "import data_generation\n",
    "import evaluation\n",
    "\n",
    "import feature_selection_hosp\n",
    "from feature_selection_hosp import *\n",
    "\n",
    "# import train\n",
    "# from train import *\n",
    "\n",
    "\n",
    "import ml_models\n",
    "from ml_models import *\n",
    "\n",
    "import dl_train\n",
    "from dl_train import *\n",
    "\n",
    "import tokenization\n",
    "from tokenization import *\n",
    "\n",
    "\n",
    "import behrt_train\n",
    "from behrt_train import *\n",
    "\n",
    "import feature_selection_icu\n",
    "from feature_selection_icu import *\n",
    "import fairness\n",
    "import callibrate_output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "nutritional-chicago",
   "metadata": {},
   "outputs": [],
   "source": [
    "importlib.reload(day_intervals_cohort)\n",
    "import day_intervals_cohort\n",
    "from day_intervals_cohort import *\n",
    "\n",
    "importlib.reload(day_intervals_cohort_v2)\n",
    "import day_intervals_cohort_v2\n",
    "from day_intervals_cohort_v2 import *\n",
    "\n",
    "importlib.reload(data_generation_icu)\n",
    "import data_generation_icu\n",
    "importlib.reload(data_generation)\n",
    "import data_generation\n",
    "\n",
    "importlib.reload(feature_selection_hosp)\n",
    "import feature_selection_hosp\n",
    "from feature_selection_hosp import *\n",
    "\n",
    "importlib.reload(feature_selection_icu)\n",
    "import feature_selection_icu\n",
    "from feature_selection_icu import *\n",
    "\n",
    "importlib.reload(tokenization)\n",
    "import tokenization\n",
    "from tokenization import *\n",
    "\n",
    "importlib.reload(ml_models)\n",
    "import ml_models\n",
    "from ml_models import *\n",
    "\n",
    "importlib.reload(dl_train)\n",
    "import dl_train\n",
    "from dl_train import *\n",
    "\n",
    "importlib.reload(behrt_train)\n",
    "import behrt_train\n",
    "from behrt_train import *\n",
    "\n",
    "importlib.reload(fairness)\n",
    "import fairness\n",
    "\n",
    "importlib.reload(callibrate_output)\n",
    "import callibrate_output\n",
    "\n",
    "importlib.reload(evaluation)\n",
    "import evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "eligible-demand",
   "metadata": {},
   "source": [
    "# Welcome to your MIMIC-IV Project"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "jewish-crown",
   "metadata": {},
   "source": [
    "This repository explains the steps to download and clean MIMIC-IV dataset for analysis.\n",
    "The repository is compatible with MIMIC-IV v1.0 and MIMIC-IV v2.0\n",
    "\n",
    "Please go to:\n",
    "- https://physionet.org/content/mimiciv/1.0/ for v1.0\n",
    "- https://physionet.org/content/mimiciv/2.0/ for v2.0\n",
    "\n",
    "Follow instructions to get access to MIMIC-IV dataset.\n",
    "\n",
    "Download the files using your terminal: \n",
    "- wget -r -N -c -np --user mehakg --ask-password https://physionet.org/files/mimiciv/1.0/ or\n",
    "- wget -r -N -c -np --user mehakg --ask-password https://physionet.org/files/mimiciv/2.0/\n",
    "        \n",
    "\n",
    "Save downloaded files in the parent directory of this github repo. \n",
    "\n",
    "The structure should look like below for v1.0-\n",
    "- mimiciv/1.0/core\n",
    "- mimiciv/1.0/hosp\n",
    "- mimiciv/1.0/icu\n",
    "\n",
    "The structure should look like below for v2.0-\n",
    "- mimiciv/2.0/hosp\n",
    "- mimiciv/2.0/icu"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "happy-tissue",
   "metadata": {},
   "source": [
    "## 1. DATA EXTRACTION\n",
    "Please run below cell to select option for cohort selection.\n",
    "The cohort will be svaed in **./data/cohort/**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "structured-dimension",
   "metadata": {
    "tags": [
     "{",
     "\"tags\":",
     "[",
     "\"hide-input\"",
     "]",
     "}"
    ]
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Please select the approriate version of MIMIC-IV for which you have downloaded data ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5913860dc40c43e0851efc6c478acbb6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Version 1', 'Version 2'), value='Version 2')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Please select what prediction task you want to perform ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e3c3922286044fa091597ea5610390a9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('Mortality', 'Length of Stay', 'Readmission', 'Phenotype'), value='Mortality')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"Please select the approriate version of MIMIC-IV for which you have downloaded data ?\")\n",
    "version = widgets.RadioButtons(options=['Version 1','Version 2'],value='Version 2')\n",
    "display(version)\n",
    "\n",
    "print(\"Please select what prediction task you want to perform ?\")\n",
    "radio_input4 = widgets.RadioButtons(options=['Mortality','Length of Stay','Readmission','Phenotype'],value='Mortality')\n",
    "display(radio_input4)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "written-syndicate",
   "metadata": {},
   "source": [
    "### Refining Cohort and Prediction Task Definition\n",
    "\n",
    "Based on your current selection following block will provide option to further refine prediction task and cohort associated with it:\n",
    "\n",
    "- First you will refine the prediction task choosing from following options -\n",
    "    - **length of Stay** - You can select from two predefined options or enter custom number of days to predict length os stay greater than number of days.\n",
    "\n",
    "    - **Readmission** - You can select from two predefined options or enter custom number of days to predict readmission after \"number of days\" after previous admission.\n",
    "\n",
    "    - **Phenotype Prediction** - You can select from four major chronic diseases to predict its future outcome\n",
    "\n",
    "        - Heart failure\n",
    "        - CAD (Coronary Artery Disease)\n",
    "        - CKD (Chronic Kidney Disease)\n",
    "        - COPD (Chronic obstructive pulmonary disease)\n",
    "\n",
    "- Second, you will choode whether to perfom above task using ICU or non-ICU admissions data\n",
    "\n",
    "- Third, you can refine the refine the cohort selection for any of the above choosen prediction tasks by including the admission samples admitted with particular chronic disease - \n",
    "    - Heart failure\n",
    "    - CAD (Coronary Artery Disease)\n",
    "    - CKD (Chronic Kidney Disease)\n",
    "    - COPD (Chronic obstructive pulmonary disease)\n",
    "    \n",
    "print(\"**Please run below cell to extract the cohort for selected options**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "broke-spirituality",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extract Data\n",
      "Please select below if you want to work with ICU or Non-ICU data ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a78a64d20d3b45d494427f706da5ad92",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('ICU', 'Non-ICU'), value='ICU')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Please select if you want to perform choosen prediction task for a specific disease.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cb78ac9ee74f4854af700e5da2033ffd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('No Disease Filter', 'Heart Failure', 'CKD', 'CAD', 'COPD'), value='No Disease Filter')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "if radio_input4.value=='Length of Stay':\n",
    "    radio_input2 = widgets.RadioButtons(options=['Length of Stay ge 3','Length of Stay ge 7','Custom'],value='Length of Stay ge 3')\n",
    "    display(radio_input2)\n",
    "    text1=widgets.IntSlider(\n",
    "    value=3,\n",
    "    min=1,\n",
    "    max=10,\n",
    "    step=1,\n",
    "    disabled=False,\n",
    "    continuous_update=False,\n",
    "    orientation='horizontal',\n",
    "    readout=True,\n",
    "    readout_format='d'\n",
    ")\n",
    "    display(widgets.HBox([widgets.Label('Length of stay ge (in days)',layout={'width': '180px'}), text1]))\n",
    "elif radio_input4.value=='Readmission':\n",
    "    radio_input2 = widgets.RadioButtons(options=['30 Day Readmission','60 Day Readmission','90 Day Readmission','120 Day Readmission','Custom'],value='30 Day Readmission')\n",
    "    display(radio_input2)\n",
    "    text1=widgets.IntSlider(\n",
    "    value=30,\n",
    "    min=10,\n",
    "    max=150,\n",
    "    step=10,\n",
    "    disabled=False\n",
    "    )\n",
    "    display(widgets.HBox([widgets.Label('Readmission after (in days)',layout={'width': '180px'}), text1]))\n",
    "elif radio_input4.value=='Phenotype':\n",
    "    radio_input2 = widgets.RadioButtons(options=['Heart Failure in 30 days','CAD in 30 days','CKD in 30 days','COPD in 30 days'],value='Heart Failure in 30 days')\n",
    "    display(radio_input2)\n",
    "elif radio_input4.value=='Mortality':\n",
    "    radio_input2 = widgets.RadioButtons(options=['Mortality'],value='Mortality')\n",
    "    #display(radio_input2)\n",
    "\n",
    "print(\"Extract Data\")\n",
    "print(\"Please select below if you want to work with ICU or Non-ICU data ?\")\n",
    "radio_input1 = widgets.RadioButtons(options=['ICU', 'Non-ICU'],value='ICU')\n",
    "display(radio_input1)\n",
    "\n",
    "print(\"Please select if you want to perform choosen prediction task for a specific disease.\")\n",
    "radio_input3 = widgets.RadioButtons(options=['No Disease Filter','Heart Failure','CKD','CAD','COPD'],value='No Disease Filter')\n",
    "display(radio_input3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "republican-freight",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "===========MIMIC-IV v2.0============\n",
      "EXTRACTING FOR: | ICU | MORTALITY | 0 |\n",
      "[ MORTALITY LABELS FINISHED ]\n",
      "[ COHORT SUCCESSFULLY SAVED ]\n",
      "[ SUMMARY SUCCESSFULLY SAVED ]\n",
      "Mortality FOR ICU DATA\n",
      "# Admission Records: 76943\n",
      "# Patients: 53569\n",
      "# Positive cases: 5156\n",
      "# Negative cases: 71787\n"
     ]
    }
   ],
   "source": [
    "disease_label=\"\"\n",
    "time=0\n",
    "label=radio_input4.value\n",
    "\n",
    "if label=='Readmission':\n",
    "    if radio_input2.value=='Custom':\n",
    "        time=text1.value\n",
    "    else:\n",
    "        time=int(radio_input2.value.split()[0])\n",
    "elif label=='Length of Stay':\n",
    "    if radio_input2.value=='Custom':\n",
    "        time=text1.value\n",
    "    else:\n",
    "        time=int(radio_input2.value.split()[4])\n",
    "\n",
    "if label=='Phenotype':    \n",
    "    if radio_input2.value=='Heart Failure in 30 days':\n",
    "        label='Readmission'\n",
    "        time=30\n",
    "        disease_label='I50'\n",
    "    elif radio_input2.value=='CAD in 30 days':\n",
    "        label='Readmission'\n",
    "        time=30\n",
    "        disease_label='I25'\n",
    "    elif radio_input2.value=='CKD in 30 days':\n",
    "        label='Readmission'\n",
    "        time=30\n",
    "        disease_label='N18'\n",
    "    elif radio_input2.value=='COPD in 30 days':\n",
    "        label='Readmission'\n",
    "        time=30\n",
    "        disease_label='J44'\n",
    "    \n",
    "data_icu=radio_input1.value==\"ICU\"\n",
    "data_mort=label==\"Mortality\"\n",
    "data_admn=label=='Readmission'\n",
    "data_los=label=='Length of Stay'\n",
    "        \n",
    "\n",
    "if (radio_input3.value==\"Heart Failure\"):\n",
    "    icd_code='I50'\n",
    "elif (radio_input3.value==\"CKD\"):\n",
    "    icd_code='N18'\n",
    "elif (radio_input3.value==\"COPD\"):\n",
    "    icd_code='J44'\n",
    "elif (radio_input3.value==\"CAD\"):\n",
    "    icd_code='I25'\n",
    "else:\n",
    "    icd_code='No Disease Filter'\n",
    "\n",
    "if version.value=='Version 1':\n",
    "    version_path=\"mimiciv/1.0\"\n",
    "    cohort_output = day_intervals_cohort.extract_data(radio_input1.value,label,time,icd_code, root_dir,disease_label)\n",
    "elif version.value=='Version 2':\n",
    "    version_path=\"mimiciv/2.0\"\n",
    "    cohort_output = day_intervals_cohort_v2.extract_data(radio_input1.value,label,time,icd_code, root_dir,disease_label)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "interstate-stadium",
   "metadata": {},
   "source": [
    "## 2. FEATURE SELECTION\n",
    "Features available for ICU data -\n",
    "- Diagnosis (https://mimic.mit.edu/docs/iv/modules/hosp/diagnoses_icd/)\n",
    "- Procedures (https://mimic.mit.edu/docs/iv/modules/icu/procedureevents/)\n",
    "- Medications (https://mimic.mit.edu/docs/iv/modules/icu/inputevents/)\n",
    "- Output Events (https://mimic.mit.edu/docs/iv/modules/icu/outputevents/)\n",
    "- Chart Events (https://mimic.mit.edu/docs/iv/modules/icu/chartevents/)\n",
    "\n",
    "Features available for ICU data -\n",
    "- Diagnosis (https://mimic.mit.edu/docs/iv/modules/hosp/diagnoses_icd/)\n",
    "- Procedures (https://mimic.mit.edu/docs/iv/modules/hosp/procedures_icd/)\n",
    "- Medications (https://mimic.mit.edu/docs/iv/modules/hosp/prescriptions/)\n",
    "- Lab Events (https://mimic.mit.edu/docs/iv/modules/hosp/labevents/)\n",
    "\n",
    "All features will be saved in **./data/features/**\n",
    "\n",
    "**Please run below cell to select features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "raised-olympus",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Feature Selection\n",
      "Which Features you want to include for cohort?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3a05f0536f6b4094ad9bd5830607e641",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Diagnosis')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "46756cb8b7254fa2b3778811564c6f34",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Output Events')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5b952fd269d74334bf432fe8a3521d08",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Chart Events(Labs and Vitals)')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5eb83da49678466487a79d0e94ff0e22",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Procedures')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "221998fe981440ed8f88410efcdc5ca7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "Checkbox(value=False, description='Medications')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**Please run below cell to extract selected features**\n"
     ]
    }
   ],
   "source": [
    "print(\"Feature Selection\")\n",
    "if data_icu:\n",
    "    print(\"Which Features you want to include for cohort?\")\n",
    "    check_input1 = widgets.Checkbox(description='Diagnosis')\n",
    "    display(check_input1)\n",
    "    check_input2 = widgets.Checkbox(description='Output Events')\n",
    "    display(check_input2)\n",
    "    check_input3 = widgets.Checkbox(description='Chart Events(Labs and Vitals)')\n",
    "    display(check_input3)\n",
    "    check_input4 = widgets.Checkbox(description='Procedures')\n",
    "    display(check_input4)\n",
    "    check_input5 = widgets.Checkbox(description='Medications')\n",
    "    display(check_input5)\n",
    "else:\n",
    "    print(\"Which Features you want to include for cohort?\")\n",
    "    check_input1 = widgets.Checkbox(description='Diagnosis')\n",
    "    display(check_input1)\n",
    "    check_input2 = widgets.Checkbox(description='Labs')\n",
    "    display(check_input2)\n",
    "    check_input3 = widgets.Checkbox(description='Procedures')\n",
    "    display(check_input3)\n",
    "    check_input4 = widgets.Checkbox(description='Medications')\n",
    "    display(check_input4)\n",
    "print(\"**Please run below cell to extract selected features**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "native-covering",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[EXTRACTING DIAGNOSIS DATA]\n",
      "# unique ICD-9 codes 6686\n",
      "# unique ICD-10 codes 10136\n",
      "# unique ICD-10 codes (After converting ICD-9 to ICD-10) 10428\n",
      "# unique ICD-10 codes (After clinical gruping ICD-10 codes) 1522\n",
      "# Admissions:   76921\n",
      "Total rows 1363733\n",
      "[SUCCESSFULLY SAVED DIAGNOSIS DATA]\n",
      "[EXTRACTING OUPTPUT EVENTS DATA]\n",
      "# Unique Events:   71\n",
      "# Admissions:   74770\n",
      "Total rows 4450049\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[7], line 7\u001b[0m\n\u001b[1;32m      5\u001b[0m     proc_flag\u001b[38;5;241m=\u001b[39mcheck_input4\u001b[38;5;241m.\u001b[39mvalue\n\u001b[1;32m      6\u001b[0m     med_flag\u001b[38;5;241m=\u001b[39mcheck_input5\u001b[38;5;241m.\u001b[39mvalue\n\u001b[0;32m----> 7\u001b[0m     \u001b[43mfeature_icu\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcohort_output\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mversion_path\u001b[49m\u001b[43m,\u001b[49m\u001b[43mdiag_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mout_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mchart_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mproc_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mmed_flag\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m      8\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[1;32m      9\u001b[0m     diag_flag\u001b[38;5;241m=\u001b[39mcheck_input1\u001b[38;5;241m.\u001b[39mvalue\n",
      "File \u001b[0;32m/scai_data3/scratch/fuchs_jo/CODE/MIMIC-IV-Data-Pipeline/preprocessing/hosp_module_preproc/feature_selection_icu.py:42\u001b[0m, in \u001b[0;36mfeature_icu\u001b[0;34m(cohort_output, version_path, diag_flag, out_flag, chart_flag, proc_flag, med_flag)\u001b[0m\n\u001b[1;32m     40\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m[EXTRACTING OUPTPUT EVENTS DATA]\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m     41\u001b[0m     out \u001b[38;5;241m=\u001b[39m preproc_out(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m./\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;241m+\u001b[39mversion_path\u001b[38;5;241m+\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m/icu/outputevents.csv.gz\u001b[39m\u001b[38;5;124m\"\u001b[39m, \u001b[38;5;124m'\u001b[39m\u001b[38;5;124m./data/cohort/\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;241m+\u001b[39mcohort_output\u001b[38;5;241m+\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m.csv.gz\u001b[39m\u001b[38;5;124m'\u001b[39m, \u001b[38;5;124m'\u001b[39m\u001b[38;5;124mcharttime\u001b[39m\u001b[38;5;124m'\u001b[39m, dtypes\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m, usecols\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m)\n\u001b[0;32m---> 42\u001b[0m     \u001b[43mout\u001b[49m\u001b[43m[\u001b[49m\u001b[43m[\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43msubject_id\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mhadm_id\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mstay_id\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mitemid\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mcharttime\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mintime\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mevent_time_from_admit\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m]\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mto_csv\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43m./data/features/preproc_out_icu.csv.gz\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mcompression\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mgzip\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mindex\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43;01mFalse\u001b[39;49;00m\u001b[43m)\u001b[49m\n\u001b[1;32m     43\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m[SUCCESSFULLY SAVED OUPTPUT EVENTS DATA]\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[1;32m     45\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m chart_flag:\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/core/generic.py:3204\u001b[0m, in \u001b[0;36mNDFrame.to_csv\u001b[0;34m(self, path_or_buf, sep, na_rep, float_format, columns, header, index, index_label, mode, encoding, compression, quoting, quotechar, line_terminator, chunksize, date_format, doublequote, escapechar, decimal)\u001b[0m\n\u001b[1;32m   3180\u001b[0m \u001b[38;5;28;01mfrom\u001b[39;00m \u001b[38;5;21;01mpandas\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mio\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mformats\u001b[39;00m\u001b[38;5;21;01m.\u001b[39;00m\u001b[38;5;21;01mcsvs\u001b[39;00m \u001b[38;5;28;01mimport\u001b[39;00m CSVFormatter\n\u001b[1;32m   3182\u001b[0m formatter \u001b[38;5;241m=\u001b[39m CSVFormatter(\n\u001b[1;32m   3183\u001b[0m     df,\n\u001b[1;32m   3184\u001b[0m     path_or_buf,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   3202\u001b[0m     decimal\u001b[38;5;241m=\u001b[39mdecimal,\n\u001b[1;32m   3203\u001b[0m )\n\u001b[0;32m-> 3204\u001b[0m \u001b[43mformatter\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43msave\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   3206\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m path_or_buf \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m:\n\u001b[1;32m   3207\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m formatter\u001b[38;5;241m.\u001b[39mpath_or_buf\u001b[38;5;241m.\u001b[39mgetvalue()\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/formats/csvs.py:204\u001b[0m, in \u001b[0;36mCSVFormatter.save\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    192\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[1;32m    193\u001b[0m     \u001b[38;5;66;03m# Note: self.encoding is irrelevant here\u001b[39;00m\n\u001b[1;32m    194\u001b[0m     \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mwriter \u001b[38;5;241m=\u001b[39m csvlib\u001b[38;5;241m.\u001b[39mwriter(\n\u001b[1;32m    195\u001b[0m         f,\n\u001b[1;32m    196\u001b[0m         lineterminator\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mline_terminator,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    201\u001b[0m         quotechar\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mquotechar,\n\u001b[1;32m    202\u001b[0m     )\n\u001b[0;32m--> 204\u001b[0m     \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_save\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    206\u001b[0m \u001b[38;5;28;01mfinally\u001b[39;00m:\n\u001b[1;32m    207\u001b[0m     \u001b[38;5;28;01mif\u001b[39;00m is_zip:\n\u001b[1;32m    208\u001b[0m         \u001b[38;5;66;03m# GH17778 handles zip compression separately.\u001b[39;00m\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/formats/csvs.py:325\u001b[0m, in \u001b[0;36mCSVFormatter._save\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    322\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m start_i \u001b[38;5;241m>\u001b[39m\u001b[38;5;241m=\u001b[39m end_i:\n\u001b[1;32m    323\u001b[0m     \u001b[38;5;28;01mbreak\u001b[39;00m\n\u001b[0;32m--> 325\u001b[0m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_save_chunk\u001b[49m\u001b[43m(\u001b[49m\u001b[43mstart_i\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mend_i\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/formats/csvs.py:334\u001b[0m, in \u001b[0;36mCSVFormatter._save_chunk\u001b[0;34m(self, start_i, end_i)\u001b[0m\n\u001b[1;32m    332\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m i \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mrange\u001b[39m(\u001b[38;5;28mlen\u001b[39m(\u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mblocks)):\n\u001b[1;32m    333\u001b[0m     b \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mblocks[i]\n\u001b[0;32m--> 334\u001b[0m     d \u001b[38;5;241m=\u001b[39m \u001b[43mb\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mto_native_types\u001b[49m\u001b[43m(\u001b[49m\n\u001b[1;32m    335\u001b[0m \u001b[43m        \u001b[49m\u001b[43mslicer\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mslicer\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    336\u001b[0m \u001b[43m        \u001b[49m\u001b[43mna_rep\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mna_rep\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    337\u001b[0m \u001b[43m        \u001b[49m\u001b[43mfloat_format\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfloat_format\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    338\u001b[0m \u001b[43m        \u001b[49m\u001b[43mdecimal\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdecimal\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    339\u001b[0m \u001b[43m        \u001b[49m\u001b[43mdate_format\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mdate_format\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    340\u001b[0m \u001b[43m        \u001b[49m\u001b[43mquoting\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mquoting\u001b[49m\u001b[43m,\u001b[49m\n\u001b[1;32m    341\u001b[0m \u001b[43m    \u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    343\u001b[0m     \u001b[38;5;28;01mfor\u001b[39;00m col_loc, col \u001b[38;5;129;01min\u001b[39;00m \u001b[38;5;28mzip\u001b[39m(b\u001b[38;5;241m.\u001b[39mmgr_locs, d):\n\u001b[1;32m    344\u001b[0m         \u001b[38;5;66;03m# self.data is a preallocated list\u001b[39;00m\n\u001b[1;32m    345\u001b[0m         \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdata[col_loc] \u001b[38;5;241m=\u001b[39m col\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/core/internals/blocks.py:2547\u001b[0m, in \u001b[0;36mTimeDeltaBlock.to_native_types\u001b[0;34m(self, slicer, na_rep, quoting, **kwargs)\u001b[0m\n\u001b[1;32m   2540\u001b[0m imask \u001b[38;5;241m=\u001b[39m (\u001b[38;5;241m~\u001b[39mmask)\u001b[38;5;241m.\u001b[39mravel()\n\u001b[1;32m   2542\u001b[0m \u001b[38;5;66;03m# FIXME:\u001b[39;00m\n\u001b[1;32m   2543\u001b[0m \u001b[38;5;66;03m# should use the formats.format.Timedelta64Formatter here\u001b[39;00m\n\u001b[1;32m   2544\u001b[0m \u001b[38;5;66;03m# to figure what format to pass to the Timedelta\u001b[39;00m\n\u001b[1;32m   2545\u001b[0m \u001b[38;5;66;03m# e.g. to not show the decimals say\u001b[39;00m\n\u001b[1;32m   2546\u001b[0m rvalues\u001b[38;5;241m.\u001b[39mflat[imask] \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39marray(\n\u001b[0;32m-> 2547\u001b[0m     [Timedelta(val)\u001b[38;5;241m.\u001b[39m_repr_base(\u001b[38;5;28mformat\u001b[39m\u001b[38;5;241m=\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mall\u001b[39m\u001b[38;5;124m\"\u001b[39m) \u001b[38;5;28;01mfor\u001b[39;00m val \u001b[38;5;129;01min\u001b[39;00m values\u001b[38;5;241m.\u001b[39mravel()[imask]],\n\u001b[1;32m   2548\u001b[0m     dtype\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mobject\u001b[39m,\n\u001b[1;32m   2549\u001b[0m )\n\u001b[1;32m   2550\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m rvalues\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/core/internals/blocks.py:2547\u001b[0m, in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m   2540\u001b[0m imask \u001b[38;5;241m=\u001b[39m (\u001b[38;5;241m~\u001b[39mmask)\u001b[38;5;241m.\u001b[39mravel()\n\u001b[1;32m   2542\u001b[0m \u001b[38;5;66;03m# FIXME:\u001b[39;00m\n\u001b[1;32m   2543\u001b[0m \u001b[38;5;66;03m# should use the formats.format.Timedelta64Formatter here\u001b[39;00m\n\u001b[1;32m   2544\u001b[0m \u001b[38;5;66;03m# to figure what format to pass to the Timedelta\u001b[39;00m\n\u001b[1;32m   2545\u001b[0m \u001b[38;5;66;03m# e.g. to not show the decimals say\u001b[39;00m\n\u001b[1;32m   2546\u001b[0m rvalues\u001b[38;5;241m.\u001b[39mflat[imask] \u001b[38;5;241m=\u001b[39m np\u001b[38;5;241m.\u001b[39marray(\n\u001b[0;32m-> 2547\u001b[0m     [\u001b[43mTimedelta\u001b[49m\u001b[43m(\u001b[49m\u001b[43mval\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_repr_base\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;28;43mformat\u001b[39;49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[38;5;124;43mall\u001b[39;49m\u001b[38;5;124;43m\"\u001b[39;49m\u001b[43m)\u001b[49m \u001b[38;5;28;01mfor\u001b[39;00m val \u001b[38;5;129;01min\u001b[39;00m values\u001b[38;5;241m.\u001b[39mravel()[imask]],\n\u001b[1;32m   2548\u001b[0m     dtype\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mobject\u001b[39m,\n\u001b[1;32m   2549\u001b[0m )\n\u001b[1;32m   2550\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m rvalues\n",
      "File \u001b[0;32mpandas/_libs/tslibs/timedeltas.pyx:1085\u001b[0m, in \u001b[0;36mpandas._libs.tslibs.timedeltas._Timedelta._repr_base\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "if data_icu:\n",
    "    diag_flag=check_input1.value\n",
    "    out_flag=check_input2.value\n",
    "    chart_flag=check_input3.value\n",
    "    proc_flag=check_input4.value\n",
    "    med_flag=check_input5.value\n",
    "    feature_icu(cohort_output, version_path,diag_flag,out_flag,chart_flag,proc_flag,med_flag)\n",
    "else:\n",
    "    diag_flag=check_input1.value\n",
    "    lab_flag=check_input2.value\n",
    "    proc_flag=check_input3.value\n",
    "    med_flag=check_input4.value\n",
    "    feature_nonicu(cohort_output, version_path,diag_flag,lab_flag,proc_flag,med_flag)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "aboriginal-upset",
   "metadata": {},
   "source": [
    "## 3. CLINICAL GROUPING\n",
    "Below you will have option to clinically group diagnosis and medications.\n",
    "Grouping medical codes will reduce dimensional space of features.\n",
    "\n",
    "Default options selected below will group medical codes to reduce feature dimension space.\n",
    "\n",
    "**Please run below cell to select preprocessing for diferent features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "partial-manhattan",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to group ICD 10 DIAG codes ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40b30e0ca36f4414b91b2e528c39902c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=2, layout=Layout(width='100%'), options=('Keep both ICD-9 and ICD-10 codes', 'Convert ICD-9…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**Please run below cell to perform feature preprocessing**\n"
     ]
    }
   ],
   "source": [
    "if data_icu:\n",
    "    if diag_flag:\n",
    "        print(\"Do you want to group ICD 10 DIAG codes ?\")\n",
    "        radio_input4 = widgets.RadioButtons(options=['Keep both ICD-9 and ICD-10 codes','Convert ICD-9 to ICD-10 codes','Convert ICD-9 to ICD-10 and group ICD-10 codes'],value='Convert ICD-9 to ICD-10 and group ICD-10 codes',layout={'width': '100%'})\n",
    "        display(radio_input4)   \n",
    "    \n",
    "else:\n",
    "    if diag_flag:\n",
    "        print(\"Do you want to group ICD 10 DIAG codes ?\")\n",
    "        radio_input4 = widgets.RadioButtons(options=['Keep both ICD-9 and ICD-10 codes','Convert ICD-9 to ICD-10 codes','Convert ICD-9 to ICD-10 and group ICD-10 codes'],value='Convert ICD-9 to ICD-10 and group ICD-10 codes',layout={'width': '100%'})\n",
    "        display(radio_input4)     \n",
    "    if med_flag:\n",
    "        print(\"Do you want to group Medication codes to use Non propietary names?\")\n",
    "        radio_input5 = widgets.RadioButtons(options=['Yes','No'],value='Yes',layout={'width': '100%'})\n",
    "        display(radio_input5)\n",
    "    if proc_flag:\n",
    "        print(\"Which ICD codes for Procedures you want to keep in data?\")\n",
    "        radio_input6 = widgets.RadioButtons(options=['ICD-9 and ICD-10','ICD-10'],value='ICD-10',layout={'width': '100%'})\n",
    "        display(radio_input6)\n",
    "print(\"**Please run below cell to perform feature preprocessing**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "descending-symphony",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[PROCESSING DIAGNOSIS DATA]\n",
      "Total number of rows 1291431\n",
      "[SUCCESSFULLY SAVED DIAGNOSIS DATA]\n"
     ]
    }
   ],
   "source": [
    "group_diag=False\n",
    "group_med=False\n",
    "group_proc=False\n",
    "if data_icu:\n",
    "    if diag_flag:\n",
    "        group_diag=radio_input4.value\n",
    "    preprocess_features_icu(cohort_output, diag_flag, group_diag,False,False,False,0,0)\n",
    "else:\n",
    "    if diag_flag:\n",
    "        group_diag=radio_input4.value\n",
    "    if med_flag:\n",
    "        group_med=radio_input5.value\n",
    "    if proc_flag:\n",
    "        group_proc=radio_input6.value\n",
    "    preprocess_features_hosp(cohort_output, diag_flag,proc_flag,med_flag,False,group_diag,group_med,group_proc,False,False,0,0)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "immediate-captain",
   "metadata": {},
   "source": [
    "### 4. SUMMARY OF FEATURES\n",
    "\n",
    "This step will generate summary of all features extracted so far.<br>\n",
    "It will save summary files in **./data/summary/**<br>\n",
    "- These files provide summary about **mean frequency** of medical codes per admission.<br>\n",
    "- It also provides **total occurrence count** of each medical code.<br>\n",
    "- For labs and chart events it will also provide <br>**missing %** which tells how many rows for a certain medical code has missing value.\n",
    "\n",
    "Please use this information to further refine your cohort by selecting <br>which medical codes in each feature you want to keep and <br>which codes you would like to remove for downstream analysis tasks.\n",
    "\n",
    "**Please run below cell to generate summary files**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "thick-residence",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[GENERATING FEATURE SUMMARY]\n",
      "[SUCCESSFULLY SAVED FEATURE SUMMARY]\n"
     ]
    }
   ],
   "source": [
    "if data_icu:\n",
    "    generate_summary_icu(diag_flag,proc_flag,med_flag,out_flag,chart_flag)\n",
    "else:\n",
    "    generate_summary_hosp(diag_flag,proc_flag,med_flag,lab_flag)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "northern-architecture",
   "metadata": {},
   "source": [
    "## 5. Feature Selection\n",
    "\n",
    "based on the files generated in previous step and other infromation gathered by you,<br>\n",
    "Please select which medical codes you want to include in this study.\n",
    "\n",
    "Please run below cell to to select options for which features you want to perform feature selection.\n",
    "\n",
    "- Select **Yes** if you want to select a subset of medical codes for that feature and<br> **edit** the corresponding feature file for it.\n",
    "- Select **No** if you want to keep all the codes in a feature."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "immediate-seafood",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do Feature Selection for Diagnosis \n",
      " (If yes, please edit list of codes in ./data/summary/diag_features.csv)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b3d9e6dd125f4affae0715d724a0d735",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Yes', 'No'), value='No')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do Feature Selection for Medication \n",
      " (If yes, please edit list of codes in ./data/summary/med_features.csv)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "eec1b8b5ead74d61abb0d96ff8425a6d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Yes', 'No'), value='No')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do Feature Selection for Procedures \n",
      " (If yes, please edit list of codes in ./data/summary/proc_features.csv)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "58b743cf98824398985e69e0d962839d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Yes', 'No'), value='No')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do Feature Selection for Output event \n",
      " (If yes, please edit list of codes in ./data/summary/out_features.csv)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0b68269dab5642959f904a4da6ec6062",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Yes', 'No'), value='No')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do Feature Selection for Chart events \n",
      " (If yes, please edit list of codes in ./data/summary/chart_features.csv)\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a168adc651d34059a30f322fac23238c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('Yes', 'No'), value='No')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**Please run below cell to perform feature selection**\n"
     ]
    }
   ],
   "source": [
    "if data_icu:\n",
    "    if diag_flag:\n",
    "        print(\"Do you want to do Feature Selection for Diagnosis \\n (If yes, please edit list of codes in ./data/summary/diag_features.csv)\")\n",
    "        radio_input4 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input4)       \n",
    "    if med_flag:\n",
    "        print(\"Do you want to do Feature Selection for Medication \\n (If yes, please edit list of codes in ./data/summary/med_features.csv)\")\n",
    "        radio_input5 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input5)   \n",
    "    if proc_flag:\n",
    "        print(\"Do you want to do Feature Selection for Procedures \\n (If yes, please edit list of codes in ./data/summary/proc_features.csv)\")\n",
    "        radio_input6 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input6)   \n",
    "    if out_flag:\n",
    "        print(\"Do you want to do Feature Selection for Output event \\n (If yes, please edit list of codes in ./data/summary/out_features.csv)\")\n",
    "        radio_input7 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input7)  \n",
    "    if chart_flag:\n",
    "        print(\"Do you want to do Feature Selection for Chart events \\n (If yes, please edit list of codes in ./data/summary/chart_features.csv)\")\n",
    "        radio_input8 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input8)  \n",
    "else:\n",
    "    if diag_flag:\n",
    "        print(\"Do you want to do Feature Selection for Diagnosis \\n (If yes, please edit list of codes in ./data/summary/diag_features.csv)\")\n",
    "        radio_input4 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input4)         \n",
    "    if med_flag:\n",
    "        print(\"Do you want to do Feature Selection for Medication \\n (If yes, please edit list of codes in ./data/summary/med_features.csv)\")\n",
    "        radio_input5 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input5)   \n",
    "    if proc_flag:\n",
    "        print(\"Do you want to do Feature Selection for Procedures \\n (If yes, please edit list of codes in ./data/summary/proc_features.csv)\")\n",
    "        radio_input6 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input6)   \n",
    "    if lab_flag:\n",
    "        print(\"Do you want to do Feature Selection for Labs \\n (If yes, please edit list of codes in ./data/summary/lab_features.csv)\")\n",
    "        radio_input7 = widgets.RadioButtons(options=['Yes','No'],value='No')\n",
    "        display(radio_input7)   \n",
    "print(\"**Please run below cell to perform feature selection**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "perceived-python",
   "metadata": {},
   "outputs": [],
   "source": [
    "select_diag=False\n",
    "select_med=False\n",
    "select_proc=False\n",
    "select_lab=False\n",
    "select_out=False\n",
    "select_chart=False\n",
    "\n",
    "if data_icu:\n",
    "    if diag_flag:\n",
    "        select_diag=radio_input4.value == 'Yes'\n",
    "    if med_flag:\n",
    "        select_med=radio_input5.value == 'Yes'\n",
    "    if proc_flag:\n",
    "        select_proc=radio_input6.value == 'Yes'\n",
    "    if out_flag:\n",
    "        select_out=radio_input7.value == 'Yes'\n",
    "    if chart_flag:\n",
    "        select_chart=radio_input8.value == 'Yes'\n",
    "    features_selection_icu(cohort_output, diag_flag,proc_flag,med_flag,out_flag, chart_flag,select_diag,select_med,select_proc,select_out,select_chart)\n",
    "else:\n",
    "    if diag_flag:\n",
    "        select_diag=radio_input4.value == 'Yes'\n",
    "    if med_flag:\n",
    "        select_med=radio_input5.value == 'Yes'\n",
    "    if proc_flag:\n",
    "        select_proc=radio_input6.value == 'Yes'\n",
    "    if lab_flag:\n",
    "        select_lab=radio_input7.value == 'Yes'\n",
    "    features_selection_hosp(cohort_output, diag_flag,proc_flag,med_flag,lab_flag,select_diag,select_med,select_proc,select_lab)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "comfortable-director",
   "metadata": {},
   "source": [
    "## 6. CLEANING OF FEATURES\n",
    "Below you will have option to to clean lab and chart events by performing outlier removal and unit conversion.\n",
    "\n",
    "Outlier removal is performed to remove values higher than selected **right threshold** percentile and lower than selected **left threshold** percentile among all values for each itemid. \n",
    "\n",
    "**Please run below cell to select preprocessing for diferent features**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "moderate-forum",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Outlier removal in values of chart events ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "230feeb2fdb042f89da391c03297a761",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(layout=Layout(height='40px', width='100%'), options=('No outlier detection', 'Impute Outlier (def…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d4b55fceb20d41dea26a2ad0cb1a3b43",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Label(value='Right Outlier Threshold', layout=Layout(width='150px')), IntSlider(value=98, layou…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0d5c42cd1a504bc3811f52cc448f2553",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Label(value='Left Outlier Threshold', layout=Layout(width='150px')), IntSlider(value=0, layout=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**Please run below cell to perform feature preprocessing**\n"
     ]
    }
   ],
   "source": [
    "if data_icu:\n",
    "    if chart_flag:\n",
    "        print(\"Outlier removal in values of chart events ?\")\n",
    "        layout = widgets.Layout(width='100%', height='40px') #set width and height\n",
    "\n",
    "        radio_input5 = widgets.RadioButtons(options=['No outlier detection','Impute Outlier (default:98)','Remove outliers (default:98)'],value='No outlier detection',layout=layout)\n",
    "        display(radio_input5)\n",
    "        outlier=widgets.IntSlider(\n",
    "        value=98,\n",
    "        min=90,\n",
    "        max=99,\n",
    "        step=1,\n",
    "        disabled=False,layout={'width': '100%'}\n",
    "        )\n",
    "        left_outlier=widgets.IntSlider(\n",
    "        value=0,\n",
    "        min=0,\n",
    "        max=10,\n",
    "        step=1,\n",
    "        disabled=False,layout={'width': '100%'}\n",
    "        )\n",
    "        #display(oulier)\n",
    "        display(widgets.HBox([widgets.Label('Right Outlier Threshold',layout={'width': '150px'}), outlier]))\n",
    "        display(widgets.HBox([widgets.Label('Left Outlier Threshold',layout={'width': '150px'}), left_outlier]))\n",
    "    \n",
    "else:      \n",
    "    if lab_flag:\n",
    "        print(\"Outlier removal in values of lab events ?\")\n",
    "        layout = widgets.Layout(width='100%', height='40px') #set width and height\n",
    "\n",
    "        radio_input7 = widgets.RadioButtons(options=['No outlier detection','Impute Outlier (default:98)','Remove outliers (default:98)'],value='No outlier detection',layout=layout)\n",
    "        display(radio_input7)\n",
    "        outlier=widgets.IntSlider(\n",
    "        value=98,\n",
    "        min=90,\n",
    "        max=99,\n",
    "        step=1,\n",
    "        disabled=False,layout={'width': '100%'}\n",
    "        )\n",
    "        left_outlier=widgets.IntSlider(\n",
    "        value=0,\n",
    "        min=0,\n",
    "        max=10,\n",
    "        step=1,\n",
    "        disabled=False,layout={'width': '100%'}\n",
    "        )\n",
    "        #display(oulier)\n",
    "        display(widgets.HBox([widgets.Label('Right Outlier Threshold',layout={'width': '150px'}), outlier]))\n",
    "        display(widgets.HBox([widgets.Label('Left Outlier Threshold',layout={'width': '150px'}), left_outlier]))\n",
    "print(\"**Please run below cell to perform feature preprocessing**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "impossible-mailman",
   "metadata": {},
   "outputs": [],
   "source": [
    "thresh=0\n",
    "if data_icu:\n",
    "    if chart_flag:\n",
    "        clean_chart=radio_input5.value!='No outlier detection'\n",
    "        impute_outlier_chart=radio_input5.value=='Impute Outlier (default:98)'\n",
    "        thresh=outlier.value\n",
    "        left_thresh=left_outlier.value\n",
    "    preprocess_features_icu(cohort_output, False, False,chart_flag,clean_chart,impute_outlier_chart,thresh,left_thresh)\n",
    "else:\n",
    "    if lab_flag:\n",
    "        clean_lab=radio_input7.value!='No outlier detection'\n",
    "        impute_outlier=radio_input7.value=='Impute Outlier (default:98)'\n",
    "        thresh=outlier.value\n",
    "        left_thresh=left_outlier.value\n",
    "    preprocess_features_hosp(cohort_output, False,False,False,lab_flag,False,False,False,clean_lab,impute_outlier,thresh,left_thresh)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "independent-academy",
   "metadata": {},
   "source": [
    "## 7. Time-Series Representation\n",
    "In this section, please choose how you want to process and represent time-series data.\n",
    "\n",
    "- First option is to select the length of time-series data you want to include for this study. (Default is 72 hours)\n",
    "\n",
    "- Second option is to select bucket size which tells in what size time windows you want to divide your time-series.<br>\n",
    "For example, if you select **2** bucket size, it wil aggregate data for every 2 hours and <br>a time-series of length 24 hours will be represented as time-series with 12 time-windows <br>where data for every 2 hours is agggregated from original raw time-series.\n",
    "\n",
    "During this step, we will also save the time-series data in data dictionaries in the format that can be directly used for following deep learning analysis.\n",
    "\n",
    "### Imputation\n",
    "You can also choose if you want to impute lab/chart values. The imputation will be done by froward fill and mean or median imputation.<br>\n",
    "Values will be forward fill first and if no value exists for that admission we will use mean or median value for the patient.\n",
    "\n",
    "The data dictionaries will be saved in **./data/dict/**\n",
    "\n",
    "Please refer the readme to know the structure of data dictionaries.\n",
    "\n",
    "**Please run below cell to select time-series representation**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "mechanical-three",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=======Time-series Data Represenation=======\n",
      "Length of data to be included for time-series prediction ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5ee8c4a8256340bfb4848e3c00f4066c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('First 72 hours', 'First 48 hours', 'First 24 hours', 'Custom'), value='First 72 hours')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "86338e4c3a824fa19ccc6f0f11ef8295",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Label(value='Fisrt (in hours):', layout=Layout(width='150px')), IntSlider(value=72, description…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "What time bucket size you want to choose ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c81192e3dec74f5ca2a73d617ad2e908",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('1 hour', '2 hour', '3 hour', '4 hour', '5 hour', 'Custom'), value='1 hour')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bbfa08d38d4247e39728d2c864aae495",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Label(value='Bucket Size (in hours):', layout=Layout(width='150px')), IntSlider(value=1, max=6,…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to forward fill and mean or median impute lab/chart values to form continuous data signal?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1e86f904e4bb425cb42390f008e38704",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('No Imputation', 'forward fill and mean', 'forward fill and median'), value='fo…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "If you have choosen mortality prediction task, then what prediction window length you want to keep?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "84b0c757a3494f428459467b1d855705",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('2 hours', '4 hours', '6 hours', '8 hours', 'Custom'), value='2 hours')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e0841b7ef9954ceb9595bde13a97b656",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(Label(value='Prediction window (in hours)', layout=Layout(width='180px')), IntSlider(value=2, m…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "**Please run below cell to perform time-series represenation and save in data dictionaries**\n"
     ]
    }
   ],
   "source": [
    "print(\"=======Time-series Data Represenation=======\")\n",
    "\n",
    "print(\"Length of data to be included for time-series prediction ?\")\n",
    "if(data_mort):\n",
    "    radio_input8 = widgets.RadioButtons(options=['First 72 hours','First 48 hours','First 24 hours','Custom'],value='First 72 hours')\n",
    "    display(radio_input8)\n",
    "    text2=widgets.IntSlider(\n",
    "    value=72,\n",
    "    min=24,\n",
    "    max=72,\n",
    "    step=1,\n",
    "    description='Fisrt',\n",
    "    disabled=False\n",
    "    )\n",
    "    display(widgets.HBox([widgets.Label('Fisrt (in hours):',layout={'width': '150px'}), text2]))\n",
    "elif(data_admn):\n",
    "    radio_input8 = widgets.RadioButtons(options=['Last 72 hours','Last 48 hours','Last 24 hours','Custom'],value='Last 72 hours')\n",
    "    display(radio_input8)\n",
    "    text2=widgets.IntSlider(\n",
    "    value=72,\n",
    "    min=24,\n",
    "    max=72,\n",
    "    step=1,\n",
    "    description='Last',\n",
    "    disabled=False\n",
    "    )\n",
    "    display(widgets.HBox([widgets.Label('Last (in hours):',layout={'width': '150px'}), text2]))\n",
    "elif(data_los):\n",
    "    radio_input8 = widgets.RadioButtons(options=['First 12 hours','First 24 hours','Custom'],value='First 24 hours')\n",
    "    display(radio_input8)\n",
    "    text2=widgets.IntSlider(\n",
    "    value=72,\n",
    "    min=12,\n",
    "    max=72,\n",
    "    step=1,\n",
    "    description='First',\n",
    "    disabled=False\n",
    "    )\n",
    "    display(widgets.HBox([widgets.Label('Fisrt (in hours):',layout={'width': '150px'}), text2]))\n",
    "    \n",
    "    \n",
    "print(\"What time bucket size you want to choose ?\")\n",
    "radio_input7 = widgets.RadioButtons(options=['1 hour','2 hour','3 hour','4 hour','5 hour','Custom'],value='1 hour')\n",
    "display(radio_input7)\n",
    "text1=widgets.IntSlider(\n",
    "    value=1,\n",
    "    min=1,\n",
    "    max=6,\n",
    "    step=1,\n",
    "    disabled=False\n",
    "    )\n",
    "#display(text1)\n",
    "display(widgets.HBox([widgets.Label('Bucket Size (in hours):',layout={'width': '150px'}), text1]))\n",
    "print(\"Do you want to forward fill and mean or median impute lab/chart values to form continuous data signal?\")\n",
    "radio_impute = widgets.RadioButtons(options=['No Imputation', 'forward fill and mean','forward fill and median'],value='forward fill and mean') # default: No Imputation\n",
    "display(radio_impute)   \n",
    "\n",
    "radio_input6 = widgets.RadioButtons(options=['0 hours','2 hours','4 hours','6 hours'],value='0 hours')\n",
    "if(data_mort):\n",
    "    print(\"If you have choosen mortality prediction task, then what prediction window length you want to keep?\")\n",
    "    radio_input6 = widgets.RadioButtons(options=['2 hours','4 hours','6 hours','8 hours','Custom'],value='2 hours')\n",
    "    display(radio_input6)\n",
    "    text3=widgets.IntSlider(\n",
    "    value=2,\n",
    "    min=2,\n",
    "    max=8,\n",
    "    step=1,\n",
    "    disabled=False\n",
    "    )\n",
    "    display(widgets.HBox([widgets.Label('Prediction window (in hours)',layout={'width': '180px'}), text3]))\n",
    "print(\"**Please run below cell to perform time-series represenation and save in data dictionaries**\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "indie-appendix",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "XYZ\n",
      "Cohort Output: cohort_icu_mortality_0_\n",
      "\n",
      "Mortality Indicator (if_mort): True\n",
      "\n",
      "Admission Indicator (if_admn): False\n",
      "\n",
      "Length of Stay Indicator (if_los): False\n",
      "\n",
      "Feature Condition (feat_cond): True\n",
      "\n",
      "Feature Procedure (feat_proc): True\n",
      "\n",
      "Feature Outcome (feat_out): True\n",
      "\n",
      "Feature Chart (feat_chart): True\n",
      "\n",
      "Feature Medication (feat_med): True\n",
      "\n",
      "Imputation Strategy (impute): Mean\n",
      "\n",
      "Include Time: 72\n",
      "\n",
      "Bucket Size: 1\n",
      "\n",
      "Prediction Window (predW): 2\n",
      "\n"
     ]
    },
    {
     "ename": "NameError",
     "evalue": "name 'STOP' is not defined",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[16], line 22\u001b[0m\n\u001b[1;32m     20\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m data_icu:\n\u001b[1;32m     21\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mXYZ\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m---> 22\u001b[0m     gen\u001b[38;5;241m=\u001b[39m\u001b[43mdata_generation_icu\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mGenerator\u001b[49m\u001b[43m(\u001b[49m\u001b[43mcohort_output\u001b[49m\u001b[43m,\u001b[49m\u001b[43mdata_mort\u001b[49m\u001b[43m,\u001b[49m\u001b[43mdata_admn\u001b[49m\u001b[43m,\u001b[49m\u001b[43mdata_los\u001b[49m\u001b[43m,\u001b[49m\u001b[43mdiag_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mproc_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mout_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mchart_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mmed_flag\u001b[49m\u001b[43m,\u001b[49m\u001b[43mimpute\u001b[49m\u001b[43m,\u001b[49m\u001b[43minclude\u001b[49m\u001b[43m,\u001b[49m\u001b[43mbucket\u001b[49m\u001b[43m,\u001b[49m\u001b[43mpredW\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m     23\u001b[0m     \u001b[38;5;66;03m#gen=data_generation_icu.Generator(cohort_output,data_mort,diag_flag,False,False,chart_flag,False,impute,include,bucket,predW)\u001b[39;00m\n\u001b[1;32m     24\u001b[0m     \u001b[38;5;66;03m#if chart_flag:\u001b[39;00m\n\u001b[1;32m     25\u001b[0m     \u001b[38;5;66;03m#    gen=data_generation_icu.Generator(cohort_output,data_mort,False,False,False,chart_flag,False,impute,include,bucket,predW)\u001b[39;00m\n\u001b[1;32m     26\u001b[0m \u001b[38;5;28;01melse\u001b[39;00m:\n\u001b[1;32m     27\u001b[0m     gen\u001b[38;5;241m=\u001b[39mdata_generation\u001b[38;5;241m.\u001b[39mGenerator(cohort_output,data_mort,data_admn,data_los,diag_flag,lab_flag,proc_flag,med_flag,impute,include,bucket,predW)\n",
      "File \u001b[0;32m/scai_data3/scratch/fuchs_jo/CODE/MIMIC-IV-Data-Pipeline/model/data_generation_icu.py:44\u001b[0m, in \u001b[0;36mGenerator.__init__\u001b[0;34m(self, cohort_output, if_mort, if_admn, if_los, feat_cond, feat_proc, feat_out, feat_chart, feat_med, impute, include_time, bucket, predW)\u001b[0m\n\u001b[1;32m     40\u001b[0m \u001b[38;5;28;01mfor\u001b[39;00m key, value \u001b[38;5;129;01min\u001b[39;00m data_dict\u001b[38;5;241m.\u001b[39mitems():\n\u001b[1;32m     41\u001b[0m     \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124mf\u001b[39m\u001b[38;5;124m\"\u001b[39m\u001b[38;5;132;01m{\u001b[39;00mkey\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;124m: \u001b[39m\u001b[38;5;132;01m{\u001b[39;00mvalue\u001b[38;5;132;01m}\u001b[39;00m\u001b[38;5;130;01m\\n\u001b[39;00m\u001b[38;5;124m\"\u001b[39m)\n\u001b[0;32m---> 44\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[43mSTOP\u001b[49m)     \n\u001b[1;32m     46\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mdata \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mgenerate_adm()\n\u001b[1;32m     47\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124m[ READ COHORT ]\u001b[39m\u001b[38;5;124m\"\u001b[39m)\n",
      "\u001b[0;31mNameError\u001b[0m: name 'STOP' is not defined"
     ]
    }
   ],
   "source": [
    "if (radio_input6.value=='Custom'):\n",
    "    predW=int(text3.value)\n",
    "else:\n",
    "    predW=int(radio_input6.value[0].strip())\n",
    "if (radio_input7.value=='Custom'):\n",
    "    bucket=int(text1.value)\n",
    "else:\n",
    "    bucket=int(radio_input7.value[0].strip())\n",
    "if (radio_input8.value=='Custom'):\n",
    "    include=int(text2.value)\n",
    "else:\n",
    "    include=int(radio_input8.value.split()[1])\n",
    "if (radio_impute.value=='forward fill and mean'):\n",
    "    impute='Mean'\n",
    "elif (radio_impute.value=='forward fill and median'):\n",
    "    impute='Median'\n",
    "else:\n",
    "    impute=False\n",
    "\n",
    "\n",
    "if data_icu:\n",
    "    gen=data_generation_icu.Generator(cohort_output,data_mort,data_admn,data_los,diag_flag,proc_flag,out_flag,chart_flag,med_flag,impute,include,bucket,predW)\n",
    "    #gen=data_generation_icu.Generator(cohort_output,data_mort,diag_flag,False,False,chart_flag,False,impute,include,bucket,predW)\n",
    "    #if chart_flag:\n",
    "    #    gen=data_generation_icu.Generator(cohort_output,data_mort,False,False,False,chart_flag,False,impute,include,bucket,predW)\n",
    "else:\n",
    "    gen=data_generation.Generator(cohort_output,data_mort,data_admn,data_los,diag_flag,lab_flag,proc_flag,med_flag,impute,include,bucket,predW)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "lined-reset",
   "metadata": {},
   "source": [
    "## 8. Machine Learning Models\n",
    "\n",
    "Below we provide options to select -\n",
    "- Type of machine learning model\n",
    "- Wheteher to concatenate or aggregate time-series features.\n",
    "    For example, if the EHR data has collected value for Blood Pressure for one year over 4 time windows of 3 months each then,\n",
    "    - **Conactenate** will concatenate all four values resulting in 4 different features for blood pressure,\n",
    "    - **Aggregate** will aggreagte(mean) over four tiem windows resulting in one feature for blood pressure."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "consolidated-former",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "=======Machine :earning Models=======\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0eb17e0ad4334762aa04c9e45680f251",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('Logistic Regression', 'Random Forest', 'Gradient Bossting', 'Xgboost'), value='Logistic…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you wnat to conactenate the time-series feature\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "02f701360fe24ac7a4fd8e69ce4b91ae",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('Conactenate', 'Aggregate'), value='Conactenate')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Please select below option for cross-validation\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40ac7b2d942941be87be804a8d021759",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('No CV', '5-fold CV', '10-fold CV'), value='5-fold CV')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do oversampling for minority calss ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "080aa1b32ca84edd9f88aa60a29f2fcd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('True', 'False'), value='True')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(\"=======Machine :earning Models=======\")\n",
    "radio_input5 = widgets.RadioButtons(options=['Logistic Regression','Random Forest','Gradient Bossting','Xgboost'],value='Logistic Regression')\n",
    "display(radio_input5)\n",
    "print(\"Do you wnat to conactenate the time-series feature\")\n",
    "radio_input6 = widgets.RadioButtons(options=['Conactenate','Aggregate'],value='Conactenate')\n",
    "display(radio_input6)\n",
    "print(\"Please select below option for cross-validation\")\n",
    "radio_input7 = widgets.RadioButtons(options=['No CV','5-fold CV','10-fold CV'],value='5-fold CV')\n",
    "display(radio_input7)\n",
    "print(\"Do you want to do oversampling for minority calss ?\")\n",
    "radio_input8 = widgets.RadioButtons(options=['True','False'],value='True')\n",
    "display(radio_input8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "f2fddbec",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "promising-miller",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Total Samples 23472\n",
      "Positive Samples 2798\n",
      "=============OVERSAMPLING===============\n",
      "Total Samples 41348\n",
      "Positive Samples 20674\n",
      "=================== 0 FOLD=====================\n",
      "train_hids 33076\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "Cell \u001b[0;32mIn[13], line 7\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[38;5;28;01melif\u001b[39;00m radio_input7\u001b[38;5;241m.\u001b[39mvalue\u001b[38;5;241m==\u001b[39m\u001b[38;5;124m'\u001b[39m\u001b[38;5;124m10-fold CV\u001b[39m\u001b[38;5;124m'\u001b[39m:\n\u001b[1;32m      6\u001b[0m     cv\u001b[38;5;241m=\u001b[39m\u001b[38;5;28mint\u001b[39m(\u001b[38;5;241m10\u001b[39m)\n\u001b[0;32m----> 7\u001b[0m ml\u001b[38;5;241m=\u001b[39m\u001b[43mml_models\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mML_models\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdata_icu\u001b[49m\u001b[43m,\u001b[49m\u001b[43mcv\u001b[49m\u001b[43m,\u001b[49m\u001b[43mradio_input5\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mvalue\u001b[49m\u001b[43m,\u001b[49m\u001b[43mconcat\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mradio_input6\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mvalue\u001b[49m\u001b[38;5;241;43m==\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mConactenate\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43moversampling\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43mradio_input8\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mvalue\u001b[49m\u001b[38;5;241;43m==\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43mTrue\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m/scai_data3/scratch/fuchs_jo/CODE/MIMIC-IV-Data-Pipeline/model/ml_models.py:42\u001b[0m, in \u001b[0;36mML_models.__init__\u001b[0;34m(self, data_icu, k_fold, model_type, concat, oversampling)\u001b[0m\n\u001b[1;32m     40\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39moversampling\u001b[38;5;241m=\u001b[39moversampling\n\u001b[1;32m     41\u001b[0m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mloss\u001b[38;5;241m=\u001b[39mevaluation\u001b[38;5;241m.\u001b[39mLoss(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mcpu\u001b[39m\u001b[38;5;124m'\u001b[39m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m,\u001b[38;5;28;01mTrue\u001b[39;00m)\n\u001b[0;32m---> 42\u001b[0m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mml_train\u001b[49m\u001b[43m(\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m/scai_data3/scratch/fuchs_jo/CODE/MIMIC-IV-Data-Pipeline/model/ml_models.py:103\u001b[0m, in \u001b[0;36mML_models.ml_train\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    101\u001b[0m         concat_cols\u001b[38;5;241m.\u001b[39mextend(cols_t)\n\u001b[1;32m    102\u001b[0m \u001b[38;5;28mprint\u001b[39m(\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mtrain_hids\u001b[39m\u001b[38;5;124m'\u001b[39m,\u001b[38;5;28mlen\u001b[39m(train_hids))\n\u001b[0;32m--> 103\u001b[0m X_train,Y_train\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mgetXY\u001b[49m\u001b[43m(\u001b[49m\u001b[43mtrain_hids\u001b[49m\u001b[43m,\u001b[49m\u001b[43mlabels\u001b[49m\u001b[43m,\u001b[49m\u001b[43mconcat_cols\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    104\u001b[0m \u001b[38;5;66;03m#encoding categorical\u001b[39;00m\n\u001b[1;32m    105\u001b[0m gen_encoder \u001b[38;5;241m=\u001b[39m LabelEncoder()\n",
      "File \u001b[0;32m/scai_data3/scratch/fuchs_jo/CODE/MIMIC-IV-Data-Pipeline/model/ml_models.py:192\u001b[0m, in \u001b[0;36mML_models.getXY\u001b[0;34m(self, ids, labels, concat_cols)\u001b[0m\n\u001b[1;32m    189\u001b[0m     y\u001b[38;5;241m=\u001b[39mlabels[labels[\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mhadm_id\u001b[39m\u001b[38;5;124m'\u001b[39m]\u001b[38;5;241m==\u001b[39msample][\u001b[38;5;124m'\u001b[39m\u001b[38;5;124mlabel\u001b[39m\u001b[38;5;124m'\u001b[39m]\n\u001b[1;32m    191\u001b[0m \u001b[38;5;66;03m#print(sample)\u001b[39;00m\n\u001b[0;32m--> 192\u001b[0m dyn\u001b[38;5;241m=\u001b[39m\u001b[43mpd\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mread_csv\u001b[49m\u001b[43m(\u001b[49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43m./data/csv/\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;241;43m+\u001b[39;49m\u001b[38;5;28;43mstr\u001b[39;49m\u001b[43m(\u001b[49m\u001b[43msample\u001b[49m\u001b[43m)\u001b[49m\u001b[38;5;241;43m+\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[38;5;124;43m/dynamic.csv\u001b[39;49m\u001b[38;5;124;43m'\u001b[39;49m\u001b[43m,\u001b[49m\u001b[43mheader\u001b[49m\u001b[38;5;241;43m=\u001b[39;49m\u001b[43m[\u001b[49m\u001b[38;5;241;43m0\u001b[39;49m\u001b[43m,\u001b[49m\u001b[38;5;241;43m1\u001b[39;49m\u001b[43m]\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    194\u001b[0m \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39mconcat:\n\u001b[1;32m    195\u001b[0m     dyn\u001b[38;5;241m.\u001b[39mcolumns\u001b[38;5;241m=\u001b[39mdyn\u001b[38;5;241m.\u001b[39mcolumns\u001b[38;5;241m.\u001b[39mdroplevel(\u001b[38;5;241m0\u001b[39m)\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/parsers.py:676\u001b[0m, in \u001b[0;36m_make_parser_function.<locals>.parser_f\u001b[0;34m(filepath_or_buffer, sep, delimiter, header, names, index_col, usecols, squeeze, prefix, mangle_dupe_cols, dtype, engine, converters, true_values, false_values, skipinitialspace, skiprows, skipfooter, nrows, na_values, keep_default_na, na_filter, verbose, skip_blank_lines, parse_dates, infer_datetime_format, keep_date_col, date_parser, dayfirst, cache_dates, iterator, chunksize, compression, thousands, decimal, lineterminator, quotechar, quoting, doublequote, escapechar, comment, encoding, dialect, error_bad_lines, warn_bad_lines, delim_whitespace, low_memory, memory_map, float_precision)\u001b[0m\n\u001b[1;32m    623\u001b[0m     engine_specified \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;01mFalse\u001b[39;00m\n\u001b[1;32m    625\u001b[0m kwds\u001b[38;5;241m.\u001b[39mupdate(\n\u001b[1;32m    626\u001b[0m     delimiter\u001b[38;5;241m=\u001b[39mdelimiter,\n\u001b[1;32m    627\u001b[0m     engine\u001b[38;5;241m=\u001b[39mengine,\n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m    673\u001b[0m     skip_blank_lines\u001b[38;5;241m=\u001b[39mskip_blank_lines,\n\u001b[1;32m    674\u001b[0m )\n\u001b[0;32m--> 676\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[43m_read\u001b[49m\u001b[43m(\u001b[49m\u001b[43mfilepath_or_buffer\u001b[49m\u001b[43m,\u001b[49m\u001b[43m \u001b[49m\u001b[43mkwds\u001b[49m\u001b[43m)\u001b[49m\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/parsers.py:454\u001b[0m, in \u001b[0;36m_read\u001b[0;34m(filepath_or_buffer, kwds)\u001b[0m\n\u001b[1;32m    451\u001b[0m     \u001b[38;5;28;01mreturn\u001b[39;00m parser\n\u001b[1;32m    453\u001b[0m \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m--> 454\u001b[0m     data \u001b[38;5;241m=\u001b[39m \u001b[43mparser\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mread\u001b[49m\u001b[43m(\u001b[49m\u001b[43mnrows\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m    455\u001b[0m \u001b[38;5;28;01mfinally\u001b[39;00m:\n\u001b[1;32m    456\u001b[0m     parser\u001b[38;5;241m.\u001b[39mclose()\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/parsers.py:1133\u001b[0m, in \u001b[0;36mTextFileReader.read\u001b[0;34m(self, nrows)\u001b[0m\n\u001b[1;32m   1131\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mread\u001b[39m(\u001b[38;5;28mself\u001b[39m, nrows\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[1;32m   1132\u001b[0m     nrows \u001b[38;5;241m=\u001b[39m _validate_integer(\u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mnrows\u001b[39m\u001b[38;5;124m\"\u001b[39m, nrows)\n\u001b[0;32m-> 1133\u001b[0m     ret \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_engine\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mread\u001b[49m\u001b[43m(\u001b[49m\u001b[43mnrows\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   1135\u001b[0m     \u001b[38;5;66;03m# May alter columns / col_dict\u001b[39;00m\n\u001b[1;32m   1136\u001b[0m     index, columns, col_dict \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_create_index(ret)\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/io/parsers.py:2037\u001b[0m, in \u001b[0;36mCParserWrapper.read\u001b[0;34m(self, nrows)\u001b[0m\n\u001b[1;32m   2035\u001b[0m \u001b[38;5;28;01mdef\u001b[39;00m \u001b[38;5;21mread\u001b[39m(\u001b[38;5;28mself\u001b[39m, nrows\u001b[38;5;241m=\u001b[39m\u001b[38;5;28;01mNone\u001b[39;00m):\n\u001b[1;32m   2036\u001b[0m     \u001b[38;5;28;01mtry\u001b[39;00m:\n\u001b[0;32m-> 2037\u001b[0m         data \u001b[38;5;241m=\u001b[39m \u001b[38;5;28;43mself\u001b[39;49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43m_reader\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mread\u001b[49m\u001b[43m(\u001b[49m\u001b[43mnrows\u001b[49m\u001b[43m)\u001b[49m\n\u001b[1;32m   2038\u001b[0m     \u001b[38;5;28;01mexcept\u001b[39;00m \u001b[38;5;167;01mStopIteration\u001b[39;00m:\n\u001b[1;32m   2039\u001b[0m         \u001b[38;5;28;01mif\u001b[39;00m \u001b[38;5;28mself\u001b[39m\u001b[38;5;241m.\u001b[39m_first_chunk:\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:860\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader.read\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:875\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader._read_low_memory\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:952\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader._read_rows\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:1084\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader._convert_column_data\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:1130\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader._convert_tokens\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32mpandas/_libs/parsers.pyx:1186\u001b[0m, in \u001b[0;36mpandas._libs.parsers.TextReader._convert_with_dtype\u001b[0;34m()\u001b[0m\n",
      "File \u001b[0;32m~/.conda/envs/mimic-iv-38/lib/python3.8/site-packages/pandas/core/dtypes/common.py:1611\u001b[0m, in \u001b[0;36mis_extension_array_dtype\u001b[0;34m(arr_or_dtype)\u001b[0m\n\u001b[1;32m   1566\u001b[0m \u001b[38;5;250m\u001b[39m\u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1567\u001b[0m \u001b[38;5;124;03mCheck if an object is a pandas extension array type.\u001b[39;00m\n\u001b[1;32m   1568\u001b[0m \n\u001b[0;32m   (...)\u001b[0m\n\u001b[1;32m   1608\u001b[0m \u001b[38;5;124;03mFalse\u001b[39;00m\n\u001b[1;32m   1609\u001b[0m \u001b[38;5;124;03m\"\"\"\u001b[39;00m\n\u001b[1;32m   1610\u001b[0m dtype \u001b[38;5;241m=\u001b[39m \u001b[38;5;28mgetattr\u001b[39m(arr_or_dtype, \u001b[38;5;124m\"\u001b[39m\u001b[38;5;124mdtype\u001b[39m\u001b[38;5;124m\"\u001b[39m, arr_or_dtype)\n\u001b[0;32m-> 1611\u001b[0m \u001b[38;5;28;01mreturn\u001b[39;00m \u001b[38;5;28misinstance\u001b[39m(dtype, ExtensionDtype) \u001b[38;5;129;01mor\u001b[39;00m \u001b[43mregistry\u001b[49m\u001b[38;5;241;43m.\u001b[39;49m\u001b[43mfind\u001b[49m\u001b[43m(\u001b[49m\u001b[43mdtype\u001b[49m\u001b[43m)\u001b[49m \u001b[38;5;129;01mis\u001b[39;00m \u001b[38;5;129;01mnot\u001b[39;00m \u001b[38;5;28;01mNone\u001b[39;00m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "if radio_input7.value=='No CV':\n",
    "    cv=0\n",
    "elif radio_input7.value=='5-fold CV':\n",
    "    cv=int(5)\n",
    "elif radio_input7.value=='10-fold CV':\n",
    "    cv=int(10)\n",
    "ml=ml_models.ML_models(data_icu,cv,radio_input5.value,concat=radio_input6.value=='Conactenate',oversampling=radio_input8.value=='True')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ordinary-chancellor",
   "metadata": {},
   "source": [
    "## 9. Deep Learning Models\n",
    "- Time-series LSTM and Time-series CNN which will only use time-series events like medications, charts, labs, output events to train model.\n",
    "\n",
    "- Hybrid LSTM and Hybrid CNN will use static data - diagnosis, demographic data aong with other time-series data to train model.\n",
    "\n",
    "- LSTM with Attention model will use attention layer to rank the important features and learn to predict output. It will use both static and time-series data.\n",
    "\n",
    "**Go to ./model/parameter.py and define all variables needed for model building and training**\n",
    "\n",
    "**Please run below cell to select which model to use**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "operational-pride",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d162f67f625a457c8a9801aa324f5be3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('Time-series LSTM', 'Time-series CNN', 'Hybrid LSTM', 'Hybrid CNN'), value='Time-series …"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Please select below option for cross-validation\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0a23df702c704ed5a63944d897653f56",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(index=1, options=('No CV', '5-fold CV', '10-fold CV'), value='5-fold CV')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Do you want to do oversampling for minority calss ?\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f8f5b62f5a8c4641ab60a4cd150455c7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "RadioButtons(options=('True', 'False'), value='True')"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "radio_input6=widgets.RadioButtons(options=['Time-series LSTM','Time-series CNN','Hybrid LSTM','Hybrid CNN'],value='Time-series LSTM')\n",
    "display(radio_input6)\n",
    "print(\"Please select below option for cross-validation\")\n",
    "radio_input7 = widgets.RadioButtons(options=['No CV','5-fold CV','10-fold CV'],value='5-fold CV')\n",
    "display(radio_input7)\n",
    "print(\"Do you want to do oversampling for minority calss ?\")\n",
    "radio_input8 = widgets.RadioButtons(options=['True','False'],value='True')\n",
    "display(radio_input8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "golden-stewart",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "===============MODEL TRAINING===============\n",
      "Total Samples 500\n",
      "Positive Samples 233\n",
      "[ MODEL CREATED ]\n",
      "LSTMBase(\n",
      "  (med): ValEmbed(\n",
      "    (codeEmbed): BatchNorm1d(163, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (fc): Linear(in_features=163, out_features=152, bias=True)\n",
      "  )\n",
      "  (proc): CodeEmbed(\n",
      "    (codeEmbed): Embedding(157, 52)\n",
      "    (fc): Linear(in_features=8164, out_features=152, bias=True)\n",
      "  )\n",
      "  (out): CodeEmbed(\n",
      "    (codeEmbed): Embedding(70, 52)\n",
      "    (fc): Linear(in_features=3640, out_features=152, bias=True)\n",
      "  )\n",
      "  (chart): ValEmbed(\n",
      "    (codeEmbed): BatchNorm1d(446, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (fc): Linear(in_features=446, out_features=152, bias=True)\n",
      "  )\n",
      "  (cond): StatEmbed(\n",
      "    (codeEmbed): Embedding(1492, 52)\n",
      "    (fc): Linear(in_features=77584, out_features=152, bias=True)\n",
      "  )\n",
      "  (ethEmbed): Embedding(9, 152, padding_idx=0)\n",
      "  (genderEmbed): Embedding(3, 152, padding_idx=0)\n",
      "  (ageEmbed): Embedding(74, 152, padding_idx=0)\n",
      "  (insEmbed): Embedding(4, 152, padding_idx=0)\n",
      "  (embedfc): Linear(in_features=1368, out_features=152, bias=True)\n",
      "  (rnn): LSTM(152, 256, num_layers=2, batch_first=True)\n",
      "  (fc1): Linear(in_features=256, out_features=128, bias=True)\n",
      "  (fc2): Linear(in_features=128, out_features=1, bias=True)\n",
      ")\n",
      "=================== 0 FOLD=====================\n",
      "======= EPOCH 0.0 ========\n",
      "BCE Loss: 1.30\n",
      "AU-ROC: 0.68\n",
      "AU-PRC: 0.63\n",
      "AU-PRC Baaseline: 0.47\n",
      "Accuracy: 0.64\n",
      "Precision: 0.60\n",
      "Recall: 0.70\n",
      "Specificity: 0.58\n",
      "NPV: 0.69\n",
      "ECE: 0.07\n",
      "MCE: 0.17\n",
      "======= VALIDATION ========\n",
      "BCE Loss: 1.26\n",
      "AU-ROC: 0.72\n",
      "AU-PRC: 0.77\n",
      "AU-PRC Baaseline: 0.53\n",
      "Accuracy: 0.70\n",
      "Precision: 0.74\n",
      "Recall: 0.67\n",
      "Specificity: 0.74\n",
      "NPV: 0.67\n",
      "ECE: 0.14\n",
      "MCE: 0.44\n",
      "Validation results improved\n",
      "Updating Model\n",
      "======= EPOCH 1.0 ========\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[1;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[1;32m<ipython-input-159-22c6e1034467>\u001b[0m in \u001b[0;36m<module>\u001b[1;34m\u001b[0m\n\u001b[0;32m      7\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m      8\u001b[0m \u001b[1;32mif\u001b[0m \u001b[0mdata_icu\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m----> 9\u001b[1;33m     \u001b[0mmodel\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdl_train\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mDL_models\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdata_icu\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdiag_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mproc_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mout_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mchart_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmed_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mradio_input6\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalue\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mcv\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0moversampling\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mradio_input8\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalue\u001b[0m\u001b[1;33m==\u001b[0m\u001b[1;34m'True'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmodel_name\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'attn_icu_read'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mtrain\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     10\u001b[0m \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     11\u001b[0m     \u001b[0mmodel\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdl_train\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mDL_models\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdata_icu\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdiag_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mproc_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmed_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mlab_flag\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mradio_input6\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalue\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mcv\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0moversampling\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mradio_input8\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mvalue\u001b[0m\u001b[1;33m==\u001b[0m\u001b[1;34m'True'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mmodel_name\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;34m'attn_icu_read'\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mtrain\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mTrue\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Desktop\\MIMIC-IV-Data-Pipeline\\model\\dl_train.py\u001b[0m in \u001b[0;36m__init__\u001b[1;34m(self, data_icu, diag_flag, proc_flag, out_flag, chart_flag, med_flag, lab_flag, model_type, k_fold, oversampling, model_name, train)\u001b[0m\n\u001b[0;32m     81\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0mtrain\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     82\u001b[0m             \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"===============MODEL TRAINING===============\"\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m---> 83\u001b[1;33m             \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mdl_train\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m     84\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m     85\u001b[0m         \u001b[1;32melse\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Desktop\\MIMIC-IV-Data-Pipeline\\model\\dl_train.py\u001b[0m in \u001b[0;36mdl_train\u001b[1;34m(self)\u001b[0m\n\u001b[0;32m    162\u001b[0m                 \u001b[0mprint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;34m\"======= EPOCH {:.1f} ========\"\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mepoch\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    163\u001b[0m                 \u001b[1;32mfor\u001b[0m \u001b[0mnbatch\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mint\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_hids\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m/\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 164\u001b[1;33m                     \u001b[0mmeds\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mchart\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mout\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mproc\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mlab\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mstat_train\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mdemo_train\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mY_train\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mgetXY\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mtrain_hids\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mnbatch\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mnbatch\u001b[0m\u001b[1;33m+\u001b[0m\u001b[1;36m1\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m*\u001b[0m\u001b[0margs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m,\u001b[0m\u001b[0mlabels\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    165\u001b[0m \u001b[1;31m#                     print(chart.shape)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    166\u001b[0m \u001b[1;31m#                     print(meds.shape)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Desktop\\MIMIC-IV-Data-Pipeline\\model\\dl_train.py\u001b[0m in \u001b[0;36mgetXY\u001b[1;34m(self, ids, labels)\u001b[0m\n\u001b[0;32m    315\u001b[0m \u001b[1;31m#                 print(\"key\",key)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    316\u001b[0m \u001b[1;31m#                 print(\"keys[key]\",keys[key])\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m--> 317\u001b[1;33m                 \u001b[0mdyn_temp\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdyn\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkeys\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m    318\u001b[0m                 \u001b[0mdyn_temp\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mdyn_temp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mto_numpy\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m    319\u001b[0m                 \u001b[0mdyn_temp\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mtorch\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mtensor\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mdyn_temp\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\DSRA\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   2773\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mis_unique\u001b[0m \u001b[1;32mand\u001b[0m \u001b[0mkey\u001b[0m \u001b[1;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2774\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mnlevels\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m1\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2775\u001b[1;33m                     \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2776\u001b[0m                 \u001b[1;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_item_cache\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2777\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\DSRA\\lib\\site-packages\\pandas\\core\\frame.py\u001b[0m in \u001b[0;36m_getitem_multilevel\u001b[1;34m(self, key)\u001b[0m\n\u001b[0;32m   2847\u001b[0m     \u001b[1;32mdef\u001b[0m \u001b[0m_getitem_multilevel\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mself\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2848\u001b[0m         \u001b[1;31m# self.columns is a MultiIndex\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2849\u001b[1;33m         \u001b[0mloc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2850\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mloc\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mslice\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mSeries\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mndarray\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mIndex\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2851\u001b[0m             \u001b[0mnew_columns\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[1;33m[\u001b[0m\u001b[0mloc\u001b[0m\u001b[1;33m]\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\DSRA\\lib\\site-packages\\pandas\\core\\indexes\\multi.py\u001b[0m in \u001b[0;36mget_loc\u001b[1;34m(self, key, method)\u001b[0m\n\u001b[0;32m   2651\u001b[0m         \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0misinstance\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[1;33m(\u001b[0m\u001b[0mtuple\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlist\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2652\u001b[0m             \u001b[1;31m# not including list here breaks some indexing, xref #30892\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2653\u001b[1;33m             \u001b[0mloc\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0m_get_level_indexer\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mlevel\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;36m0\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2654\u001b[0m             \u001b[1;32mreturn\u001b[0m \u001b[0m_maybe_to_slice\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mloc\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2655\u001b[0m \u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;32m~\\Anaconda3\\envs\\DSRA\\lib\\site-packages\\pandas\\core\\indexes\\multi.py\u001b[0m in \u001b[0;36m_get_level_indexer\u001b[1;34m(self, key, level, indexer)\u001b[0m\n\u001b[0;32m   2922\u001b[0m             \u001b[1;32mif\u001b[0m \u001b[0mlevel\u001b[0m \u001b[1;33m>\u001b[0m \u001b[1;36m0\u001b[0m \u001b[1;32mor\u001b[0m \u001b[0mself\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0mlexsort_depth\u001b[0m \u001b[1;33m==\u001b[0m \u001b[1;36m0\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2923\u001b[0m                 \u001b[1;31m# Desired level is not sorted\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[1;32m-> 2924\u001b[1;33m                 \u001b[0mlocs\u001b[0m \u001b[1;33m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0marray\u001b[0m\u001b[1;33m(\u001b[0m\u001b[0mlevel_codes\u001b[0m \u001b[1;33m==\u001b[0m \u001b[0mcode\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mdtype\u001b[0m\u001b[1;33m=\u001b[0m\u001b[0mbool\u001b[0m\u001b[1;33m,\u001b[0m \u001b[0mcopy\u001b[0m\u001b[1;33m=\u001b[0m\u001b[1;32mFalse\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0m\u001b[0;32m   2925\u001b[0m                 \u001b[1;32mif\u001b[0m \u001b[1;32mnot\u001b[0m \u001b[0mlocs\u001b[0m\u001b[1;33m.\u001b[0m\u001b[0many\u001b[0m\u001b[1;33m(\u001b[0m\u001b[1;33m)\u001b[0m\u001b[1;33m:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n\u001b[0;32m   2926\u001b[0m                     \u001b[1;31m# The label is present in self.levels[level] but unused:\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[1;33m\u001b[0m\u001b[0m\n",
      "\u001b[1;31mKeyboardInterrupt\u001b[0m: "
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x432 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAfEAAAHaCAYAAAAQWXCIAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy8li6FKAAAgAElEQVR4nOy9e3hV1Z3//165X8idhFuKgAkYEkUuwQgKWipoQcBCC4g1fFtKO9N+O499Oq2db/vr/dupbZ/pt9NpHaVTgqDYQgcRK6CIxlGRgIgNUkmAcIeE3Dm5J+v3R7KPJ4eck3NZa6+91vm8nicP5JyT937nc9ZZ7+y914VxzkEQBEEQhH5EqTZAEARBEERoUIgTBEEQhKZQiBMEQRCEplCIEwRBEISmUIgTBEEQhKZQiBMEQRCEpkgLccbYfzHGahljlT6eZ4yx3zDGqhljHzDGZsjyQhAEQRAmIvNMfBOA+/08/wCA/IGvDQB+L9ELQRAEQRiHtBDnnJcDaPDzkmUANvN+DgJIZ4yNkeWHIAiCIExD5T3xcQDOe3x/YeAxgiAIgiACIEbhsdkQjw25BixjbAP6L7kjPj5+5qhRowAAycnJiI2NRVNTEwaeQ2pqKurq6gAAUVFRyMrKQlNTE7q7uwEAGRkZ6OzsRFtbGwBgxIgRiI6ORnNzs1sjJSUF165dAwBER0cjMzMTjY2N6OnpAQBkZmaivb0d7e3tAICUlBQwxtDS0gIASEhIQHJyMurr6wdpNDQ0oLe3FwCQlZUFl8uFjo4OAEBqaio452htbQUAJCYmIjExEQ0N/RczYmJikJGRMUhj5MiRaG1tRWdnJwAgLS0Nvb29uH79OgAgKSkJ8fHxaGxsBADExsYiPT0d9fX16OvrAwBkZ2cP+j49PR3d3d1wuVxD1jguLg5paWm4du0aOOdgjGHkyJFobm5GV1dXQBqR8D5Zv5vI96mlpcWtEYnvU11dHaKiohz/edLpfWpsbERUVJTjP0+y3qeWxkagrw+JiYno6+1F50DN4+LiEBMdjbaB3zU6OhpJiYloHTgGAKSMGIG29nb09vaCc46YhARwALy3F7ynB/FxcYiKjnbXKyY6GgmJiW6fbOB9aGtrQ++Ar6TERPT09KBr4H2Lj49HFGNoa29HbWPjNc55NrxQGeIXAHzC4/tcAJeGeiHn/CkATwHA1KlT+YcffijfXQRx5MgRzJw5U7UNo6CaiodqKp5Ir+mLv/89HszNDUvD1d2Nsqoq1HV0IBHAnKgo3HX77WIMesCWLj071OMqL6fvAvDowCj1EgDNnPPLw/1QbGysfGcRRm6YjZi4EaqpeKim4qGahodngGcnJGDNhAm40tKC+oGrE3Ygc4rZcwDeATCFMXaBMfZFxthXGGNfGXjJXwGcBlAN4GkA/xiIrnUJiRDHCy+8oNqCcVBNxUM1FQ/VNHS8A7w0Px+fyMxEX0MDDldX2xbk0i6nc87XDPM8B/BVWccnCIIgCH9Y98BDoaKublCAJw9cJU5NSMCsvDwcrq7GrLw8ZKWmirI7JCrviYdEXFzcDY91d3fjwoUL7sESRHDMmjULJ06csP24CQkJyM3NNfIWydixY1VbMA6qqXgivab1zc2ob2kJKWjnj+mfEV2cne0OcADIzMpCVmqqbUHO+k+I9WHWrFn88OHDgx47c+YMUlJSkJWVBcaGGvROOA3OOerr69Ha2oqJEyeqtkMQRASy/f/9P6Q0NQUctK7ubkQzhoSYwM5/61tahAU5W7r0COd8lvfj2q2dbk2B8KSjo4MCPAw8p0PZBWMMWVlZxl492bhxo2oLxkE1FU+k1zQ+Ls59xjzcPWzrHvgz1dXoGJh2NxT79u1z/9/zjFzWPXLtQtzXlQMKcP0w+T2z5qMS4qCaiodqGljQeg5i6+7rQ6+fK9jeNZUd5NqFuFM7/ujoaNx+++3ur3/9138F0H+//vHHH0d+fj6Kioowe/ZsvPzyywCACRMm4NZbb3X/zNe//vWwfXzhC19ATk4OioqKhnz+l7/8JRhjPq9ozJ49G9OmTUNhYSG+//3vu5/785//jMLCQkRFRcHzdsZbb72F2267DcXFxaiurgbQP4Ng0aJFPv/gigSiorT7aDkeqql4qKb9+AvaoUahJ/sZxzNUTWUGuXYD20aOHDnsa0b/cjSuuq4KO+ao5FG48s0rfl+TmJiI999//4bHv/e97+Hy5cuorKxEfHw8rl69ijfeeMP9/IEDBwL6nQJl3bp1+NrXvoZHH330hufOnz+PV155BePHjx/0eHZ2/yJA8fHxeO211zBixAh0d3fjrrvuwgMPPICSkhIUFRXhL3/5C7785S8P+tlf/epX2LFjB2pqavD73/8ev/rVr/DjH/8Y//Iv/+LYP7jsYP369aotGAfVVDxU048ZajBasAEOAAsXLgxYXwTa/RlmLefoD5EBHo5eW1sbnn76afz7v/874uPjAQCjRo3C5z73OZH2BjFv3jxkZmYO+dxjjz2GJ5544oZwtebeM8YwYsQIAP1XELq7u92vLSgowJQpU27QjI2NRXt7O9ra2hAbG4tTp07h4sWLmD9/vshfSzt2796t2oJxUE3FQzUdjGfQXmxsDDrAAeBQRUVA+qLOyLU7E+8KY16fTNrb23G7x1J73/nOd1BQUIDx48cj1c9fXPfeey+io6MBAKWlpXjssccGPb9161b84he/uOHn8vLysH379oD97dq1C+PGjcO0adNueM5aXxkAent7MXPmTFRXV+OrX/0q7rjjDr+63/nOd7BhwwYkJibimWeewTe/+U38+Mc/DtiXqVy6NOQKwkQYUE3FQzW9EStoK6qrMTIlBQACDnAAaBhYO344fVFn5NqFuFMZ6nL6Bx98MOzPDXc5fe3atVi7dm1Y3tra2vDTn/500KhJX0RHR+P9999HU1MTHnroIVRWVvq8vw4At99+Ow4ePAgAKC8vx9ixY8E5x6pVqxAbG4tf/epXsDasIQiC0IGs1FQUDwT53IkTAw7wYPRFBbl2l9PT09NVWwiYvLw8nDt3zr1DTyhs3bp10IA562vlypUBa5w6dQpnzpzBtGnTMGHCBFy4cAEzZszAlSv99/mHqml6ejruuece7NmzJ6BjcM7xk5/8BN/73vfwwx/+ED/84Q/xyCOP4De/+U3APk1i2bJlqi0YB9VUPFTTwbi6u/HfNTVo7+lxB3nlmTNBXfq+o6QkoNeJurSuXYh7Xvp1OklJSfjiF7+Ir3/96+7bAJcvX8aWLVsC1li7di3ef//9G76CuZR+6623ora2FjU1NaipqUFubi7ee+89jB49GsDHtyjq6urc98fb29vx6quv4pZbbgnoGGVlZVi8eDEyMjLQ1taGqKgoREVFubc+jDQuXLig2oJxUE3FQzX9GGsQ2wcNDdgzUJdQgrZ+iJk/vhAR5NqFuLWnrtOw7olbX48//jgA4Cc/+Qmys7MxdepUFBUVYfny5e7R4ED/PXHrZ4YaUR4sa9aswZ133omPPvoIubm5+MMf/uD39ZcuXXL/NX758mXce++97ilj9913H5YsWQIA+O///m/k5ubinXfeweLFi7Fo0SK3RltbG8rKyvCP/9i/h803vvENrFixAt/5znfwD//wD2H/Tjpy5MgR1RaMg2oqHqppP96j0BeOG+d+LtigtabaBkq4QW7kPfFRyaOETzEbDmvDem/i4uLwxBNP4IknnrjhuZqamnCt3cBzzz037Gs8jzt27Fj3z9x22204evTokD/z0EMP4aGHHhryuaSkJBw4cMD9/d13342//e1vQbgmCIJQQyDTyGSvhR6OvnYhnpycPOxrhpvTTQwmkJoSwTF79mzVFoyDaiqeSK9pd19fwNPIAg3ayZMnh+Ql1CDX7nK6iTteqSYmwMX8icDxvGVCiIFqKp5Ir+m19vag5oEHcuk7LS0tZD+hXFrXLsStgVeEOAJZQIcIjpdeekm1BeOgmoon0ms6OikJ944ZE9Q88OGCtsLPYi8i9L3RLsQJgiAIIlRcLhfa29sB9K9SOW/MmKDngcve1CQYfe1C3Fq+lBBHXFycagvG4b0+PRE+VFPxRFpNXS4XysrK8Mwzz7iDPFR8Ba2oWxSBBrl2Ie5vCVMiNKim4vG1CQIROlRT8URSTa0Ar6urQ09Pj5BtWIcK2unTp4etO5S+L7QL8bq6OtUWhoQxhs9//vPu73t6epCdne2eZw0AL7/8MmbNmoWCggLccsst+OY3vwkA+MEPfoBx48YNmmcu6t6/r61Hz507hxEjRuCXv/zlkNuS+tp69JVXXsHMmTNx6623YubMmXjttdcAAJ2dnbj//vtRVFSE3/3ud+7Xb9iwwee0NZPZuHGjagvGQTUVT6TU1DPAs7OzUVpaKmxWjneQB7K8dSj6vtAuxANi9GiAMXFfAyub+SM5ORmVlZXuSzSvvPIKxnksGFBZWYmvfe1r2LJlC06cOIHKykpMmjTJ/fxjjz02aEU2EcvL+tp61DreAw884PNnra1H582bN+jxkSNH4sUXX8Tf/vY3lJWVuf9w2bt3L2bOnIkPPvgATz31FADg2LFj6OvrE/qXKUEQRDDIDHALzyBv6egQqm3p+0K7EA9oE/urYrciDVTvgQcecI/2fO6557BmzRr3c0888QT+z//5P+5lTGNiYtwrnMnC19ajO3fuxKRJk1BYWAgAQ+777Wvr0enTp2Ps2LEAgMLCQnR0dKCzs9O9JWlPT4/7td/73vfwox/9SOSvpA00zkA8VFPxmF7Tjo4O6QFuYQX5qYYGKYPdfKFdiGdlZam24JPVq1dj27Zt6OjowAcffDBoG8/KykrMnDnT58/+27/9m/tS+r333nvD862trUNuhHL77bfjww8/vOH1vrYedblc+PnPf47vf//77sf87aLmjx07dmD69OmIj4/HfffdhytXruCOO+7At771LezatQszZ850B36ksW7dOtUWjINqKh7TaxofH4/x48dLD3CLrNRUbFi5Utqo9aHQbpUPJ88Tv+2221BTU4PnnnsOn/70p4P62ccee8x9j3woUlJSbtjq1Bf+th79/ve/j8ceewwjRoxwP9bY2IiMjIyg/B4/fhzf/va33ceIiYnBs88+C6B/k5pFixZh165d+MY3voFz587h0UcfxdKlS4M6hs7s3LkTy5cvV23DKKim4jG9powxLF68GJ2dnUhISLjh+c6BzZ9EUvXhh5g1daq0JVq90S7Enb6L2dKlS/HNb34Tr7/+Ouo9NocvLCzEkSNHbjgzDpTW1lbcfffdQz737LPPYurUqe7vPbceBeDeevTQoUN49913sX37dnzrW99CU1MToqKi0NPT496wJRAuXLiAhx56CJs3b8bNN998w/O/+93vUFpainfeeQdxcXF4/vnnceedd0ZUiNfW1qq2YBxUU/GYWFOXy4U9e/bggQceQFJSEhhjQwY4ANQ3N6O+pUVo0DY1NUlfa90T7ULc6XzhC19AWloabr31Vrz++uvux//5n/8Zn/nMZ3DXXXdh8uTJ6Ovrw69//Wt84xvfCEg3mDNxa+tRiwkTJuDw4cMYOXIk3nzzTffjP/jBDzBixAiUlpYG9suhv4EuXrwYP/vZzzB37twbnm9sbMTu3buxb98+7Nq1C1FRUWCMoUPCYA+CIAhPPAexAcCKFSv8vj4rLc2Rm5oEg3b3xIO97Gs3ubm5+Kd/+qcbHr/tttvw61//GmvWrEFBQQGKiopw+fJl9/Oe98Rvv/12KTuc+cKq6fr1693TyXxtPfrb3/4W1dXV+PGPf+z26vkHw49+9CN897vfBWMMixYtwuHDh3HrrbfiS1/6km2/jxMYrvMggodqKh6Tauo9Cv3+++8f9mfi4+KEr7w2Z84c9/9lr+wGAIxzLkVYFgUFBfzEiRODHjtx4gQKCgo+fmD0aLEj1EeNAq6YuzPa9evXB90jt5Mb3jtDePfddwcNbCTCh2oqHlNqGuo0shd//3s8mJuL+pYWYWfMH3300Q0ze0Tos6VLj3DOZ3k/rt2ZeFtb2/AvunIF4Fzcl8EBDiDs5QeJGzl27JhqC8ZBNRWPCTUVMQ9c5BnzmTNnpOp7o12IEwRBEITFe++9J2QeuJM2NQkG7Qa2qbrsazJUU/F43hcjxEA1FY8JNb3rrrsAADNmzAh7HriIwWj+bg/KGOym3Zl4dHT0kI/rdm/fSQS0Cp4ETH7PUlJSVFswDqqpeHStqcvlct9aZYzh7rvvlrYWerAkJiZK1fdGuxBvbm6+4bGEhATU19cbHQoyabFxiUALzjnq6+t9zt/Unb1796q2YBxUU/HoWFPrHvjmzZsDGyMVAuEE7XvvvSdV3xvtLqcPRW5uLi5cuODYHc6cTmtr65A7mckmISEBubm5th+XIAg98R7EJvPETfY8b1H62oV4fHz8DY/FxsZi4sSJCtyYwauvvorZs2ertmEUnjvUEWKgmopHp5rasRuZN6EE7egAdr0MR98b7S6n63oPx8l4bzdKhA/VVDxUU/HoUlMVAW4R7KVva3dIWfreaBfiKi77ms6mTZtUWzAOqql4qKbi0aGmdm4n6otggnb//v1S9b3R7nI6QRAE4Wxe/ctf0D7EqpmdXV2ob25GVloa4gPcy5xzDly/joToaIzmHK9t3uzztcHqnzh6FA8GOC7HqffItQtxX1PMiNBJSkpSbcE4qKbioZqKR1ZN269e9RmOoSxByjlHZ18fEgLo/4PRP1ZeHtDxLQIJ2qHGbYnU90a7y+mZmZmqLRjHI488otqCcVBNxUM1FY+KmgZy6djV3Y3tZ87ANbD1NGMsoAAPVD8chtO/9957pep7o12INzY2qrZgHNu3b1dtwTiopuKhmopHVU39BZWruxtlVVU43tiIly9cEK4vAn/6b731llR9b7QL8Z6eHtUWjKOhoUG1BeOgmoqHaioelTUdKqisAK/r6EB2QgIeCGMdCVVB3traKlXfG+1CnCAIgjADz6A639AwKMBL8/ORHBsrTF+nTU2G0veFdiFO98TFs2rVKtUWjINqKh6qqXicUNOs1FRMnTABz9XUCA1wT307g/zuu++Wou8L7UKc9r4WT2VlpWoLxkE1FQ/VVDxOqenZri60A0gE8ODo0cIC3MLOID/24YdS9H1BIU7g+PHjqi0YB9VUPFRT8TilpnNHjcKCsWOxZsIEfFhTo/Wl7wNHj0rR94V2IU4QBEHoj6u7e9AUsrtGj8YnMjO1v4edn5MjTX8otAtxWjtdPPPnz1dtwTiopuKhmopHVU2tUeibq6rcQW5h56Xv9s5O4fpzZs2S6t8b7UKcMabagnHExGi3cJ/joZqKh2oqHhU19ZxG5msjUbuC/Gpjo3D96Oho6f490S7EW2y81xAphLJgP+Efqql4qKbisbum3vPA/Y1CtyPIR2VkCNc/duyYW9+OINcuxAmCIAj9CCbALWQHYWJ8vNb34AENQzwhIUG1BeOYPHmyagvGQTUVD9VUPHbVtLO3N+SFXHRbsGXcuHFS9b3RLsTt3kc2EiguLlZtwTiopuKhmorHrprGRUXh5tTUkBdy0SnI8/Pzpep7o12I19fXq7ZgHFu3blVtwTiopuKhmopHVk07u7oGfc8Yw8Jx4/DFKVNCXshFlyB//fXXpep7o12IEwRBEM6mvrkZ5xsa8OfTpwfNBY8PcDtRX+gS5Hbqaxfi0WE2AuJGUgPYeJ4IDqqpeKim4pFV07SUFDxXU4MPm5rw1/PnhWo7PWiTkpKk6nujXYjTBijiWb16tWoLxkE1FQ/VVDwyaupyuXDW5XKvhV6Sni78GE4O8nnz5knV90a7EKc9hcWzbds21RaMg2oqHqqpeETX1OVyoaysDB29vchOSDBiLfRg9cvLy6Xqe6NdiPf29qq2YBy0gI54qKbioZqKR2RNrQCvq6tDQnQ0SvPzjVgLPVj9trY2qfreaBfiBEEQhPM4duwY6urqkJ2djfyMDPcodCcGrUn62oV4VlaWagvGsXbtWtUWjINqKh6qqXhE1vTOO+/Efffdh9LSUsRGDY4WpwehSP177rlHqr432oW4y+VSbcE4KioqVFswDqqpeKim4gm3pi6XC9evXwfQP4Vszpw5PhfkclLQytSvqqqSqu+NdiHe0dGh2oJxnDx5UrUF46CaiodqKp5wamrdAy8rK3MH+XA4JWhl6l+8eFGqvjfahThBEAShFs9BbIyxoLaIdkLQmqSvXYjTgg/iWbBggWoLxkE1FQ/VVDyh1NQzwLOzs1FaWhr0nhZOC0KR+tOmTZOq7412Ic65r23kiVDp6elRbcE4qKbioZqKJ9iaighwC1ODXNQ06ED9axfira2tqi0YxxtvvKHagnFQTcUjq6ajR09wXxJ28tfo0ROE/+7B1LSzs1NYgFuYGOSVlZVS9H2hXYgTBEGI5OrVswC447/6faojLi4O+fn5wgLcwsQgl6HvC+1CPDExUbUF4ygsLFRtwTiopuKhmoonmJoyxvCpT30KX/ziF4UFuIUpQXu4uhqpEtYyyfIzFoxCnEBRUZFqC8ZBNRUP1VQ8w9XU5XLhT3/6k/s2JmMM8fHxUryYEuR13d1S9H2hXYjTBijief7551VbMA6qqXiopuLxV1NrENuJEyfw8ssv2+LHhCDvrK2Vpj8U2oU4QRAEIRfvUeiLFy+27dh2Bnl7Z6dw/dSEBKn+vdEuxGNiYlRbMA7ao108VFPxUE3FM1RNRU4jCxW7gvxqY6Nw/ZSUFOn+PdEuxDMyMlRbMI6VK1eqtmAcVFPxUE3F411TJwS4hR1BPiojQ7j+3Llz3fp2BLl2IU73xMWzZcsW1RaMg2oqHqqpeLxr+re//c0RAW4hOwgT4+OF6x84cMD9fzuCXLsQF7UaDvExwWxiTwQG1VQ8VFPxeNf0jjvuwKJFixwR4Ba6DXbr9LrPLtu/1BBnjN3PGPuIMVbNGHt8iOfTGGMvMsaOMcaOM8b+l0w/BEEQxGBcLtegKWQlJSWOCXAL3YLcTn1pIc4YiwbwHwAeADAVwBrG2FSvl30VwIec82kA7gHwK8ZYnD/dkSNHSnAb2axbt061BeOgmoqHaiqez372s+7tREUuad3Z1SVMy0KXoPW1qYws/zLPxGcDqOacn+acdwHYBmCZ12s4gBTWv4/dCAANAPyuyE9rp4unvLxctQXjoJqKh2oqFpfLhaeffhp1dXWIiopCVJS4OKhvbnZ00MrUP378uFR9b2SG+DgA5z2+vzDwmCe/BVAA4BKAvwH4J855nz9R7/sNRPicPn1atQXjoJqKh2oqDmsUent7u5RBbFlpaY4OWpn6V65ckarvjcxJ10PtEu+9j+giAO8D+CSAmwG8whh7k3M+6DdjjG0AsAHon2L21FNPAQBmz56N7OxsvPTSSwCA8ePHY+HChdi4cSOA/gX7161bh507d6K2thYAsGLFClRXV+PYsWMAgDlz5iAlJQV79+4FAEyaNAnz5s3Dpk2bAABJSUl45JFHsH37dvfI+FWrVqGystL9F9f8+fMRExOD/fv3AwAmT56M4uJibN26FUD/HuirV6/Gtm3b0DLwpq1duxYVFRU4efIkgP5LMD09Pe5dhQoLC1FUVOReUSkzMxMrV67Eli1b3INR1q1bh/LycnfntmjRIrS2tuLtt98G0L+vbV5eHnbs2AEAyMnJwfLly7Fp0yZ0DVzuWr9+PVpaWtw1Xbx4Merq6nDo0CEAwMyZM5Gbm4sXXngBADB27FgsWbIEGzduRF9fH6KiorB+/Xrs3r0bly5dAgAsW7YMFy5cwJEjRyL2fWpoaHDXVNT7tG/fPpw7dy5i36dr167hqaeeEv550od49F+0DI/k5GSUlpYiJycHtbW1+MUvfoGvfvWrAvx9TEJsCv604VEcBtBVV4fYgS2k77nnHlRVVeHixYsA+tt+b2+ve+ev8ePH46abbsKbb74JoH/O9dy5c3HgwAH3CdyCBQtw/vRpNJ87hycPH8aKT30KidHROHHiBABg4sSJGDNmjPvzlZ6ejpKSEry6fz96ursBAAsXLsTRo0dx6tQp7NmzB8XFxWhubna3n7y8PORlZ+PJbduQn5ODCePGYXZxMfbt2+f+PC1cuBCHKirQUF8PALijpAT1166hpaUFe/bsweTJk5GWloaKigoAQHZ2NqZPn459+/YBANp6enAYAJqawAc+53PmzMHly5dx5swZAEBBQQESExPx3nvv+aw1k7U/N2PsTgA/4JwvGvj+OwDAOf+Zx2teAvCvnPM3B75/DcDjnPNDvnRvu+02/sEHH0jxHKmcPXsWN910k2obRkE1FY+smvYHo5x+UCzh+4yL68T69X9ATk4damuzUVZWCpdLxiA2Br5rF+pbWnC4uhqz8vL8buIRKuHq/+TZZ/Hdhx8Wql9bW4ucnJyAXhuMPlu69AjnfJb34zIvp1cAyGeMTRwYrLYawC6v15wDsAAAGGOjAEwB4PeaGU0xEw+NMxAP1VQ8VNPw6eqKw0cfTZEc4B/j9EvfMvTb29ul6nsjLcQ55z0AvgZgL4ATAP7EOT/OGPsKY+wrAy/7MYA5jLG/AdgP4Nuc82v+dK9fvy7LcsRiXXYixEE1FQ/VVAQM+/d/Ehs3flF6gFs4MWhl6luX9WXpeyN1njjn/K+c88mc85s55z8deOxJzvmTA/+/xDlfyDm/lXNexDnX6QYVQRCE40lOdmHVqueRkmIFBENXl5ztRH3htKA1SV+7FduSkpJUWzCOadOmqbZgHFRT8VBNgyc52YXS0jIUFPwdn/60PduJ+sLJQShSf+LEiVL1vdEuxGVtSB/J5OXlqbZgHFRT8VBNg8MKcGsQ2+7dS1RbckzQytQfM2aMVH1vtAvxxsZG1RaMw5raRIiDaioeqmngeAe4HYPYAsUJQStTP9yxG8H61y7ECYIgCN84OcAtVAetSfrahXhsbKxqC8YR6JxGInCopuKhmgZGYeFxRwe4hZOCUKR+enq6VH1vZK7YJgVRBSI+Zvny5aotGAfVVDxU08A4dKgYAHD8eKFjA9zCM6hkLAijQr+kpESKvi+0OxOvH1jijhCHtSQmIQ6qqXiopr5JTnYhNfXjKQG1JWAAACAASURBVGSHDs12fIBbmHZG/urAcsGi9X2hXYj39fndH4UIgS4J2wZGOlRT8VBNh8a6B15aWuYR5HphUpA3SFhZ0N8VBO1CnCAIgujHcxBbb280enujVVsKGVOCvKq2Voq+L7QL8ezsbNUWjGP9+vWqLRgH1VQ8VNPB6DAKPVhMCPINn/ucNP2h0C7EW2z8CydSsLbGI8RBNRUP1fRjTAxwCzuDvH1ge1ORnDt1Sqp/b7QL8U4JRY90rD2qCXFQTcVDNe0nNrbL2AC3sCvIrzY2Ctevq6uT7t8T7UKcIAgikunujsOHH041NsAt7AjyURkZ2l66t9AuxGmeuHgWL16s2oJxUE3FQzX9mNdfvwdPP73e2AC3kB2EifHxwvWLi4vd/7cjyLUL8e7ubtUWjKOurk61BeOgmoonkmuanOzCmjVrkJbW7H6suztOoSP70G2wW3Nz86DvZfvXLsRdLpdqC8Zx6NAh1RaMg2oqnkitqTWIbcqUKfj0p/+q2o4SdArykydPStX3RrsQJwiCiBQGj0Kvxa5dS1VbUoZOQW6nvnYhnpxs9j0gFcycOVO1BeOgmoon0mp64zSyMq3ugesUhCL1/e17L8O/diFOu5iJJzc3V7UF46CaiieSajr0PHC9biU6OWhl6meNHClV3xvtQrypqUm1BeN44YUXVFswDqqpeCKppgUFH2o/D9zJQStT/92DB6Xqe6NdiBMEQZjO4cPF2L3709oGOODsoDVJX7sQj4uLjGkVdjJ27FjVFoyDaioe02uanOwaNIXs8OFibQPcQpcgFKmfmZUlVd8b7UI8LS1NtQXjWLJkiWoLxkE1FY/JNbXuga9bt2lQkJuAE4NWpv5sj8VeZOh7o12IX7t2TbUF49i4caNqC8ZBNRWPqTX1HMTW3R2Lnp4Y1ZaE47SglakfykY94fjXLsQ556otGEdfX59qC8ZBNRWPiTU1eTcyb5wUtDL1Q22nofrXLsQZY6otGEdUlHbNwPFQTcVjWk0jKcAtnBK0MvXDaaeh+NfuUzFymDl4RPCsX79etQXjoJqKx6SaRsJ2or5wQtDK1F+4cKFUfW+0C3HvxeWJ8Nm9e7dqC8ZBNRWPSTXt7o5DZWVhxAW4heqglal/qKJCqr432oV4V1eXagvGcenSJdUWjINqKh7TalpePj8ithP1halB3lBfL1XfG+1CnCAIQkeSk114+OFnkZ7+8aqTkbKdqC9MDXIZ+r7QLsTT09NVWzCOZcuWqbZgHFRT8ehcU2sQ2+TJVRG7nagvTAra+pYW3FFSIkXfF9qFeHd3t2oLxnHhwgXVFoyDaioeXWvqPQr9hRf0/WNEFiYFefXZs1L0faFdiOu2k48OHDlyRLUF46CaikfHmkbiNLJQMSXI9737rhR9X2gX4gRBEDpAAR48JgR5fk6ONP2h0C7Ek5PpQyCa2bNnq7ZgHFRT8ehW01tu+TsFeAjYGeTtnZ3C9WfddptU/95oF+KxsbGqLRhHdna2agvGQTUVj241PXJkJl58cQkFeAjYFeRXGxuF66elpUn374l2Id7U1DT8i4igeOmll1RbMA6qqXh0qGlysmvQFLIjR2ZSgIeIHUE+KiNDuH7FwGIvdgW5diFOEAThRDy3E/UMciJ0ZAdhYny81vfgAQ1DPD4+XrUF4xg/frxqC8ZBNRWPk2vqOYitqysO3d10208Uug12877tI9u/diGe6me+HBEa4S7YT9wI1VQ8Tq0pjUKXj05BPn36dKn63mgX4nV1daotGMfGjRtVWzAOqql4nFhTCnD70CXI9+3bJ1XfG+1CnCAIwgnExHRTgPvByUFrkr52IR7OhuvE0MTFRfYmDDKgmorHaTXt6YnFsWO3UYD7QKcgFKkfM8w0aNH+Gec8bBE7mTVrFj98+LBqGwRBGAJjDEDo/WBMTDd6euwYyBaeT3thuLZlCw5XV2NWXp7ftb9Dpb6lJWz9nzz7LL778MPS9P0RrD5buvQI53yW9+PandbSPHHx7Ny5U7UF46CaiscJNU1OdmHt2q3IyGh0P2ZPgOuH08+YZekfPHhQqr432oU47WImntraWtUWjINqKh7VNbUGseXnV9N2ogHi1KCVqR/MiaYI/9qFOEEQhN14j0LfuXO5akva4MSgNUlfuxDPyMhQbcE4VqxYodqCcVBNxaOqpjSNLHycHoQi9efMmSNV3xvtQrxTwq4zkU51dbVqC8ZBNRWPippSgIvDSUErU//y5ctS9b3RLsTb2tpUWzCOY8eOqbZgHFRT8aio6eTJJynABeKUoJWpf+bMGan63mgX4gRBEHZx9Oh0vPDCUgpwgTghaE3S1y7ER4wYodqCcYRyD4fwD9VUPHbVNDnZhYyMBvf3R49OpwAXjNOCUKR+QUGBVH1vtAvx6Oho1RaMIyUlRbUF46CaiseOmn68nWjZoCAnxGNqkCcmJkrV90a7EG9ublZtwTj27t2r2oJxUE3FI7umnoPYOjvj0dVF2x7LxsQgf++996To+0K7ECcIghANjUJXh4lBLkPfF9qFeHw8/XUsmkmTJqm2YBxUU/HIqmlycjIFuGJMCdrD1dVIkLDOur+11bULcbrXKJ558+aptmAcVFPxyKhpd3c3SktLKcAdgClB7oqJkaLvC+1C/Nq1a6otGMemTZtUWzAOqql4ZNQ0NjYWR48epQB3CCYEuevSJWn6Q6FdiBMEQYjknXfewVNPfYkC3CHYGeTtElYATU1IkOrfG+1CnKaYiScpKUm1BeOgmopHVE1dLhe2bNmC+vp692O0naizsCvIrzY2CtePj4+X7t8T7UI8MzNTtQXjeOSRR1RbMA6qqXhE1NTlcqGsrAynTp3Cyy+/LMAVIQs7gnxURoZw/Xvvvdetb0eQaxfijY2Nqi0Yx/bt21VbMA6qqXjCrakV4HV1dcjOzsZDDz0kyBkhC9lBmBgfL1z/rbfecv/fjiDXLsR7enpUWzCOhgZamUo0VFPxhFNT7wAvLS1FcjLdA9cB3Qa7tba2StX3RrsQJwiCCAYKcP3RLcjt1NcuxOmeuHhWrVql2oJxUE3FE2pNq6qqKMANQJegvfvuu6Xqe6NdiLe3t6u2YByVlZWqLRgH1VQ8odb09ttvx7JlyyjAbcbJQStT/+zZs1L1vaEQJ3D8+HHVFoyDaiqeYGrqcrkGTSG7/fbbKcBtxslBK1P/3LlzUvW90S7ECYIg/GHdAy8rKxsU5IS9ODloTdLXLsRp7XTxzJ8/X7UF46CaiieQmnoOYktISEBCQoINzoih0CkIReoXFRVJ1fdGuxBnjKm2YBwxMTGqLRgH1VQ8w9WURqE7D6cGrUz9YFYVFeFfuxBvsXF3mEhh//79qi0YB9VUPP5qSgHuXJwYtDL1jx07JlXfG+1CnCAIwpOenh4KcIfjtKA1SV+7EKd7XOKZPHmyagvGQTUVj6+axsTEYMaMGRTgDsfJQShSf9y4cVL1vdEuxOkDKp7i4mLVFoyDaioefzUtKSnBhg0bqH9wOE4JWpn6+fn5UvW90S7EacqIeLZu3aragnFQTcXjWVOXy4VnnnkG165dcz9Ggwn1wAlBK1P/9ddfl6rvjXYhThBEZGMNYjt9+jRtJ6opqoPWJH2pIc4Yu58x9hFjrJox9riP19zDGHufMXacMfbGcJrBDN8nAiM1NVW1BeOgmoonNTX1hlHon/nMZ1TbIkLESUEoUj8pKUmqvjfSQpwxFg3gPwA8AGAqgDWMsaler0kH8DsASznnhQA+O5wubYAintWrV6u2YBxUU/E8+OCDNArdMEwM8nnz5knR94XMM/HZAKo556c5510AtgFY5vWahwH8hXN+DgA457XDidI+zeLZtm2bagvGQTUVi8vlwm9+8xsKcAMxLcjLy8ul6PtCZoiPA3De4/sLA495MhlABmPsdcbYEcbYo8OJ9vb2CrRIALSAjgyopmI5deoUurq6KMANxaQgvyLhRDPLz+05mcM5h1oflQ9x/JkAFgBIBPAOY+wg5/zkICHGNgDYAAAZGRl46qmnAACzZ89GdnY2XnrpJQDA+PHjsXDhQmzcuBEAEBcXh3Xr1mHnzp2ore0/yV+xYgWqq6vdq+rMmTMHKSkp2Lt3LwBg0qRJmDdvHjZt2gSg//7GI488gu3bt7uvAqxatQqVlZXuXZXmz5+PmJgY94pSkydPRnFxsXs0bWpqKlavXo1t27a5O/e1a9eioqICJ0/2/6oLFixAT08P3nijf1hAYWEhioqK8PzzzwPov42wcuVKbNmyBW1tbQCAdevWoby8HKdPnwYALFq0CK2trXj77bcBANOmTUNeXh527NgBAMjJycHy5cuxadMmdHV1AQDWr1+PlpYWd00XL16Muro6HDp0CAAwc+ZM5Obm4oUXXgAAjB07FkuWLMHGjRvR19eHqKgorF+/Hrt378alS5cAAMuWLcOFCxdw5MiRiH2fGhoa3DUV9T7t27fPvUOS09+n3/z0p4h2uTBy5Ejk5eXh4MGDbo07Zs/Gkffec7fjWTNn4tKlS7h0+TKA/ik60VFR+PtHH7nrMeGmm9B85gya//53fPvAAdxz11344IMP0NHR0e+9uBg1Z8+6fd0yZQp6+/pQVVXVX48xYzB27FgcHvhdk5KSMHPGDLw7UD9CPOXl5Zg3bx7Ky8vd7/U999yDqqoqXLx4EUB/2+/t7XVvM5ualYX/qaxEZ20tUhMSkJKSgrlz5+LAgQPo7OwE0P8ZPH78OK5cuQIAmDFjBtrb23HixAkAwMSJEzFmzBj35ys9PR0lJSU4WlGB5tZWPHn4MDZ87nM4d+oUTp06hT179qC4uBjNzc3uz3leXh6yRo7EuwPtNjMrC7OLi7Fv3z7352nhwoU4VFGBhoEZU3eUlCAzOhpv1NQAO3di1m23IS0tDRUVFQCA7OxsTJ8+Hfv27QMAxMTG4lMLFuDgwYNoamoC0P/5uXz5Ms6cOQMAKCgoQGJiIt577z2fdWace+eqGBhjdwL4Aed80cD33wEAzvnPPF7zOIAEzvkPBr7/A4A9nPM/+9KdMWMG9/cLEcHjcrnozEYwkV7T//zRj7Dy5pv9nkEMh6u7G+29vRg5sMBTR0cHEhISUN/SgsPV1ZiVlxeWvgVbuhQ3nl84EQY9fAIAA9+1K6SfFP3++tP//e7d+O7DDwvVv1hXh8rz54X7Z0uXHuGcz/J+XObl9AoA+YyxiYyxOACrAXi/qy8AuJsxFsMYSwJwB4AT/kRdLpcUs5GM9ZciIY5Ir2lWWlpYly5d3d0oq6pC2cmTuDZwtm2dVcu+NEqoxc5L3+0DZ/ciabhyxdb2KS3EOec9AL4GYC/6g/lPnPPjjLGvMMa+MvCaEwD2APgAwCEAGznnlf50rctnhDisS0iEOCK9pvFxcSF3ZFaA13V0IDEmBokD00qtS7AABbnp2BXkVxsbhetfvHjR1vYpdZ445/yvnPPJnPObOec/HXjsSc75kx6v+QXnfCrnvIhz/muZfgiCsI9QOjLPAM9OSEBpfj6SY2OF6RP6YEeQj8rI0HYwnYV2K7bRIhriWbBggWoLxkE17SeYjmy4AJ82bVpY+oR+yH5/E+Pjhet7tlM72qd2IS5rIF4k09PTo9qCcVBNPyaQjqynrw+bhzkD9zW9lILcbHSbfubdTmX71y7EW1tbVVswDmu6FCEOqulghuvIYqKiMCs7Gzl+LqFb05BC0Sf0RqcgH6qdyvSvXYgTBKEnw3VkxdnZ+NItt/i8Bx6uPqE3OgW5nfrahXhiYqJqC8ZRWFio2oJxUE2HxrMjO9/Q0H8Jvb3d/XxMlO8uafz48UHpU5CrRacgFKnvr53K8E8hTqCoqEi1BeOgmvomKzUVUydMwHM1NTjT2oqXL1wI6OduuummgPUpyNXj5KCVqT9cOxXtX7sQpw1QxGMtGUqIg2rqm+vd3XjxyhW0o3+t5XuzsgL6uTfffDPgY1CQq8fJQStTP5B2KtK/diFOEIS+XO/uHjQKfc2ECfiwpsZxHTERPk4OWpP0tQvxmBiZe7ZEJrRHu3iopjfiHeCl+fn4RGZmwB1ZSkpK0MekIFeLLkEoUj+YdirCv3YhnpGRodqCcaxcuVK1BeOgmt7ImdbWIeeBB9qRzZ07N6TjUpCrxYlBK1M/2HYarn/tQpzuiYtny5Ytqi0YB9X0Rm7NzMRnJkwYch54IB3ZgQMHQj42BblanBa0MvVDaafh+NcuxH2t2kSEjrXXLyEOqmk/ru7uQVPIbs3MDHkt9M4wd5yiIFeLk4JWpn6o7TRU/9qFOEEQemCthb7Jay64P5zSERNy0P39daK+diE+cuRI1RaMY926daotGEek17S7r8+9mUlyTAySghiQ6qsjE7WpDAW5WpwYhCL1w22nwfrXLsRp7XTxlJeXq7ZgHJFcU5fLharGxoC2E/XFUB3Z8ePHhXn01CfsR3XQytQX0U6D8a9diId7X4y4kdOnT6u2YByRWlOXy4WysjJ09PaGHOAW3h3ZlStXhHq19Ak1mBrkotppoP61C3GCIJxJb28vNm/ejLq6OiRER4cV4BaeHVlLR4cgp4P1CXWYGuQy9H0xbIgzxr7GGHPM5Oy0tDTVFoxj0aJFqi0YRyTWNDo6GrNnz0ZOTg7yMzLCDnALqyPjqal0D9tATAra+pYWzJgxQ4q+LwI5Ex8NoIIx9ifG2P2MMSbMXQjQFDPx0DgD8URqTWfOnIkNGzYg1s9uZKGQlZqKKaNH02A0QzEpyC9duyZF3xfDftI4598FkA/gDwDWAahijP1fxtjNogwGw/Xr11Uc1mjefvtt1RaMI1Jqat0Dr62tdT8WHR2Nzq4u4ce6cu4cjSo3GFOC/KX/+R9b22dAfy5zzjmAKwNfPQAyAGxnjD0h0RtBEA7GCvCamhq8/PLLg56rb27WsiMm1GJCkOfn5NjaPgO5J/51xtgRAE8AeAvArZzzfwAwE8AKyf5uICkpye5DGs+0adNUWzAO02tqBXhdXR2ys7NvWCs+Ky1NeEc2ceLEfm0KcqOxM8jbJcx2mlZQYGv7DORMfCSAz3DOF3HO/8w57wYAznkfgCVS3Q1BfHy83Yc0njyaZiMck2vqHeClpaVITk4e9Jr4uDjhHdmYMWPc/6cgNxu7gvxqY6Nw/TFjxtjaPgMJ8b8CcO86whhLYYzdAQCc8xOyjPmisbHR7kMaz44dO1RbMA5TaxpIgFuI7si8xxlQkJuNHUE+KiNDuL7VTu1qn4GE+O8BeI4mcw08RhBEhFFTUxNQgFvofo+TUIvs9zcxPl779hlIiLOBgW0A3JfRA18IWTCxguaeEh+Tk5Oj2oJxmFrTwsJCrFy5MqAAtxDVkaWnp0vVJ5yJbn8IerdT2f4DCfHTA4PbYge+/gmAsjUlfX2QidBZvny5agvGYVJNXS7XoClkhYWFAQe4hYiOrKSkRKo+4Vx0CvKh2qlM/4GE+FcAzAFwEcAFAHcA2CDURRDU19erOrSxbNq0SbUF4zClptY9cO+54KEQbkf26v79UvUJZ6NLkPtqp7L8B7LYSy3nfDXnPIdzPopz/jDnPLxPcxj09fWpOrSxdElYmCPSMaGmnoPYkpOTgz77HopwOrKe7m6p+oRYnBy0MvX9tVMZ/gOZJ57AGPsqY+x3jLH/sr6EHJ0gCEcSzCj0YNGhIybCR9f3Vzf9QC6nP4P+9dMXAXgDQC4AZQtDZ2dnqzq0saxfv161BePQuaYyA9wilI5s4cKFUvUJsegUhCL1A2mnIv0HEuJ5nPPvAXBxzssALAZwa1hHDYMW+kAKZ9++faotGIeuNfXcTlRWgFsE25EdPXpUqj4hFicHrUz9QNupKP+BhLh1gb+JMVYEIA3AhJCPGCadEpbJi3TOnTun2oJx6FrT6OholJSUICcnR2qAWwTTkdXV1UnVJ8Tj1KCVqR9MOxXhP5AQf2pgP/HvAtgF4EMAPw/paARBOBKPpSAwffp0bNiwQXqAWzixIybEofv763R9vyHOGIsC0MI5b+Scl3POJw2MUv/PUA2HC80TF8/ixYtVWzAOnWpq3QO/cuWK+7Ho6GhbPQTSkRUXF0vVJ+Th9CAUqR9KOw3Hv98QH1id7WtBO5JIdwDTTIjgCOUyJeEfXWpqBfjZs2exZ8+eQWfkdjNcR9bc3CxVn5CLk4JWpn6o7TRU/4FcTn+FMfZNxtgnGGOZ1ldILgXgcrlUHdpYDh06pNqCcehQU+9R6J/97GfBGFPqyV9HdvLkSan6hHycErQy9cNpp6H4D2QN9C8M/PtVj8c4gElB+iMiiNGjJ+Dq1bOqbQTEqFE34cqVGtU2AuLVv/wF7VevDvu6zq4u1Dc3IystDfFxcTc8393Xh6rGRnT09iIhOhqjOcdrmzcH7GM4/RNHj+LB3NyA9Tzx7Mhm5eUhKzU1JJ1A9Ql7sfv9NV1/2BDnnE8U5k4Adg22iSRmzpwpXLM/wNVdmg2Gq1fFn33KqCkAtF+9GnA41re0DNkRuLq7UVZVhY7eXmQnJKA0Px/JIWws5EsfAI6Vlwet58lQHZnIPdo99Qn7cVoQitQX0U6D8R/Iim2PDvUVtssQoV3MxJMb4hkT4Rsn1NTXpblz16+jrqMjrAD3py8Kb/2skSOl6BNqcMKlbxn6otppoP4DuSde7PF1N4AfAFgqwmQoNDU1qTq0sbzwwguqLRiHU2o6VEdQkJGBz02cGFaA+9MXiaf+K6+/LkWfUIeJQf7uwYNS9H0RyAYo/9vj60sApgO48SYYQRCOJCs1FVMnTMAbXkEeboB76tvRUVbV1tJgNAMxMchl6PsikDNxb9oA5IfsKEzihhhEQ4TH2LFjVVswDifV1NXdjRevXMHJqKhBQS4SOzqy2VOm0KhyQzElaA9XVyM6MVGKvi8CuSf+ImNs18DXbgAfAVB2rTAtLU3VoY1lyZIlqi0Yh1Nqag1iq+voQGpcHIonTLClI2uXsDzyffPn0/QwgzElyPtSUmxtn4Gcif8SwK8Gvn4GYB7n/HGprvxw7do1VYc2lo0bN6q2YBxOqKlngFuD2D6RmWlLR3a1sVG4/r59+2iet+GYEOSt58/b2j4DCfFzAN7lnL/BOX8LQD1jbIJUV35QuaKUqfT19am2YByqazpUgFv3wO3oyEZlZAjXt2pKQW42dga5jCtGI+LjbW2fgYT4nwF49ki9A48pQfWKUiYSFRXK0AjCHypr2ss5NvsIcAvZHWWihI7Ms6YU5GZjV5DLuGIUFRVla/sMpKeJ4Zx3Wd8M/F/Z6LKRgueKEsD69etVWzAOlTWNZgxzRo1CzjDzwHW7dLlw4UKp+oSz0PWKkdVO7WqfgYR4HWPMPS+cMbYMgLIb0+FugkDcyO7du1VbMA4VNfW81TQtKwsbCgqGnUamU5AfqqiQqk84Dx2vGHm2UzvaZyAh/hUA/8IYO8cYOwfg2wC+LMVNAHR1dQ3/IiIoLl26pNqCcciqaaeP9u/q7samkydxua3N/Vh0gLeedAnyhvp6qfqEM9GlfVp4t1PZ/gNZ7OUU57wEwFQAhZzzOZxzWnCYIBRQ39x8Q0dgDWI753Jhz/nzIQ3+1K2jtFufUIvu7UemfiDzxP8vYyydc36dc97KGMtgjP1EqIsgSE9PV3VoY1m2bJlqC8Yhq6ZZaWmDOgLvUeifmzQp5MGfTu/I7igpkapPOBunt08LX+1Ulv9ALqc/wDl3L1jOOW8E8GlhDoKku7tb1aGN5cKFC6otGIesmsbHxbk7gvMNDT6nkYWKkzvK+gDWiKAgdw5Oaz926ftrpzL8BxLi0YyxeOsbxlgigHg/r5eKy+VSdWhjOXLkiGoLxiGzptZa6M/V1AgNcE99J3aU1QFuG0pB7gyc1n7s0h+unYr2H0iIbwGwnzH2RcbYFwG8AqAs7CMTBBEyrqgodABIBPDg6NHCAtzC6R2lan1ieHR+f3XSD2Rg2xMAfgKgAP2D2/YAuCmso4ZBcnKyqkMby+zZs1VbMA7ZNb0lPR2fmzQJayZMwIc1NY7vaEToT548Wao+IRantR+79ANtp6L8B7qs1BX0r9q2AsACACdCPmKYxAo+4yCA7Oxs1RaMQ0ZNXS4X2jzGhNySnm7bWuhO0A9l8yMKcrU4qf3YpR9MOxXh32eIM8YmM8b+P8bYCQC/BXAeAOOc38s5/21IRxNAU1PT8C8iguKll15SbcE4RNfU5XKhrKwMVU1Ng+aCA87syGToVwyx2ItIfUIOTmk/dukH207D9e/vTPzv6D/rfpBzfhfn/N/Rv246QRA2YgV4XV0dYqOikDrE1SindWSRpk/4R/f318n6/kJ8Bfovox9gjD3NGFsAQPnuI/HxygbGG8v48eNVWzAOUTX1DPDs7GzkZ2QYsxZ6sPrh3qKgIFeL6vZjl36o7TRU/z5DnHP+35zzVQBuAfA6gMcAjGKM/Z4xttDXz8kmNTVV1aGNxXtjCSJ8RNTUO8BLS0sRO8zuaE7pyGToT58+Xao+IR+T26dFOO00FP+BjE53cc63cs6XAMgF8D6Ax0N2GSZ1dXWqDm0sGzduVG3BOMKtaV9fH5555plBAR7ozAwndGQy9Pft2ydVn7AHU9unRbjtNFj/QW16zDlv4Jz/J+f8kyE7JAhiWKKiojB37lyMGjUqqAC3UN2R6aRP2I9J7Ue1flAh7gSihrmcSARPXJyy7eGNJdSaem5ecuutt2LDhg0hr43gpI5GhH6MpAVtCDWY1j4tRLXTQP1rl4hZWVmqLRjHunXrVFswjlBqev36dfzxj3/ExYsX3Y+F+0erSR3l9OJiKfqEOkxqn5b+pxYskKLvC+1CnOaJi2fnzp2qLRhHsDW9fv06Nm/ejPPnz2Pv3r0hbSfqC1M6ymd276Z72AZiSvu09A8ePChF3xfahTjtYiae2tpa1RaMI5iaWgFuDWJbtWpVyNuJ+sKEjnJsUhINRjMU9ypRLAAAIABJREFUE9qnpX/uyhUp+r7QLsQJwiS8AzyUQWyBYmdH1t7ZKVw/NSGBRpUbjClBXlVba2v71C7EMzIyVFswjhUrVqi2YByB1NTOALewqyO72tgoXH/OnDk0PcxwTAjytYsX29o+tQvxTgl/4Uc6ge7TTAROIDW9ePEirl27ZluAW9jRkY3KyBCuf/nyZbc+Bbm56H7FqPP6dVvbp3Yh3ua1+QMRPseOHVNtwTgCqemUKVOwatUqWwPcQnZHmRgfL1z/zJkz7v9TkJuNzleMzpw5Y2v71C7ECUJnXC4XLl265P5+ypQptge4hQmXLinIzUXXK0ae+na0T+1CfMSIEaotGMecOXNUWzCOoWpqrYW+efPmQUGuEp2CtqCgQKo+4Tx0vGLk2U7taJ/ahXh0dLRqC8aRkpKi2oJxeNfUczOT1NRUpKWlhaTb2dUlwt4gdAnyxMREqfqEM9GlfVp4t1PZ/rUL8ebmZtUWjGPv3r2qLRiHZ02H2o0s1Evo9c3NWnRkMvTfe+89qfqEc9GhfVoM1U5l+tcuxAlCJ0QGOABkpaVp0ZGZqE+oRff2I0tfaogzxu5njH3EGKtmjPncvpQxVswY62WMrRxOMz4+XqxJApMmTVJtwTgmTZoU1naivoiPi9OyoxGhP3r0aKn6hFic1n7s0vfXTmX4lxbijLFoAP8B4AEAUwGsYYxN9fG6nwMI6Jou3b8Vz7x581RbMI558+YhKioKd999d8jbifpCh45Mhn5hYaFUfUIsTms/dukP105F+5d5Jj4bQDXn/DTnvAvANgDLhnjd/wawA0BAi01fu3ZNnEMCALBp0ybVFoyCc+6uaWFhYVjbifrC6R2ZDP39+/dL1SfE4rT2Y5d+IO1UpH+ZIT4OwHmP7y8MPOaGMTYOwEMAnpTogyBsw+Vy4Y9//CO6PEaRh7udqC+c3JFFgj7hH93fX130YwR68maobZi891f8NYBvc857/e3axBjbAGADAGRmZuKpp54CAMyePRvZ2dl46aWXAADjx4/HwoULsXHjRgBAXFwc1q1bh507d7p3lVqxYgWqq6vdK2rNmTMHKSkp7tHEkyZNwrx589xnUklJSXjkkUewfft2NDQ0AABWrVqFyspKHD9+HAAwf/58xMTEuP8Cmzx5MoqLi7F161YAQGpqKlavXo1t27ahZeDNWrt2LSoqKnDy5EkAwIIFC9DT04M33ngDQP8ZXFFREZ5//nlYv/fKlSuxZcsW96p169atQ3l5OU6fPg0AWLRoEVpbW/H2228DAKZNm4a8vDzs2LEDAJCTk4Ply5dj06ZN7pBZv3492tra3DVdvHgx6urqcOjQIQDAzJkzkZubixdeeAEAMHbsWCxZsgQbN25EX18foqKisH79euzevds993nZsqEuuDibLz34IEZmZmJOSQkqDh9GR0cHAGB2cTFqzp51t59bpkxBb18fqqqqAABjx4zB2LFjcfjIESA6Gknjx4PFxqLb5cK31q4FAEyfMQPvf/AButrbERsbi6kFBejo7HS/b7m5ucgeORJH338fQP8to9unTcPb77yD3t5eAMDcuXNx4sQJfPT++4i97TYUFxejubkZJ0+eREtHB3ZduoR506ahaqBNZmZlYXZxMfbt2+d+nxYuXIhDFRVoqK8HANxRUoL6a9fcS8ROnjwZaWlpqKioAABkZ2dj+vTpaD53Dk8ePoyCsWPxmSVLcPDgQfeWwHPmzMHly5fdq6kVFBQgMTERp06dwp49ezB69GgUFha6Pxvx8fG499578dZbb6G1tRUAMPX227HjtdeQjv5NToqKihAdHe3+jI4bNw75+fm4fv069uzZg6SkJMybNw/l5eXuz8I999yDqqoq917s06ZNQ29vLyorK9HS0YEd587h/pISfOhR47lz5+LAgQOimhDhRXl5OebNm4euujo8efgw8nNysPT++32+T0B/H37TTTfhzTffBDD4fbKW3F6wYAGOHz+OKwO7hU2YPBkvvf02Ejo6kJqQgIkTJ2LMmDHufjA9PR0lJSV4df9+9Azsgrlw4UIcPXrU3U49P08AkJeXh6yRI1Fx8CBaOjqw9eRJrP30p3Hk4MFhP09WO/X1edq3bx8AICY2tn/v8aYmPLltG/JzcnD/Jz855OfJ38wMJnLf4kHCjN0J4Aec80UD338HADjnP/N4zRl8HPYjAbQB2MA597kZ86xZs/jhw4eleCbE0f9HmZy2JR6Ga1u24HB1NWbl5fnd9s8Xru5ulFVVoa6jA9kJCSjNz0dybKz7+fqWlrD0LX7y7LP47sMP3/C4KH1fBKvvy6co/WDxp8+WLoUebVWvzxTftcv9ndPap0Wg7dQJ/tnSpUc457O8H5d5Ob0CQD5jbCJjLA7AagC7PF/AOZ/IOZ/AOZ8AYDuAf/QX4ADQ2Ngoy2/Esn37dtUWlBPOpa2hAvz9gSsZIvQDIRL033rrLan6hDyc0H7s0g+lnYbjX1qIc857AHwN/aPOTwD4E+f8OGPsK4yxr4Sq29PTI8oiMYB1myDSCeWD5OsM3LpUHK5+MJiuP1RNReoTclHdfuzSD7Wdhupf6jxxzvlfOeeTOec3c85/OvDYk5zzGwaycc7Xcc7plJBQSrAfpEttbbjm4xK6CP1gIX21+oR/dH9/naiv3YptmZmZqi0Yx6pVq1RbcBTBfJDy09Kw5uabbwjwu+++W4h+KJiq76+mIvQJezC1fVqE206D9a9diLe3t6u2YBzWyFDiY/x9kFzd3bjocrm/z09Lu+EM/OzZsyHri8BE/eFqGqo+YT8mtk8LEe00GP8U4oR7qhwxmKE+SNY98M1VVYOC3Jtz586FpC8S0/QDqWko+oQaTGufFqLaaaD+tQtxgrATzw/S+YYG9yC2tLg4pMfFCdU3qSOTod8yMHdftD6hDpPap2x9X2gX4rR2unjmz5+v2oKjyUpNxdQJE/BcTY3PeeDeFBUVBaVvSkcjU78rKYnuYRuIKe3T0g/msx+Mvi+0C3F/K7sRoRETI3PhPv1xdXfjxStX0A4gEcCDo0cPOwo9Ojo6qGOY1pHJ0J8+cSINRjMUE9qnpd8s4ZavvytG2oV4C32AhRPMxhKRRh/n2FJd7T4DXzNhAj6sqRm2I7CWDA0Gkzqy9oElMkVy/tQpGlVuMKa0/50HDtjaPrULcYKwkyjGMG/0aIxOTERpfj4+kZlpREcjW/9qY6OW/gm1mND+83NybG2f2oV4QkKCagvGMXnyZNUWHIfnngIFGRn40i23uC+hB9IRjBs3bsjHA8GEjmxURoZwfaumFORmo/sVo4Kbb7a1fWoX4qL3ZSaA4uJi1RYchau7G/918iTOXb/ufizKayzGcB1Nfn5+WB50D/LE+Hjh+p41pSA3G52vGOXn59vaPrUL8fqBbd8IcVhbphIfzwO/4HLhlYsX4W+XP38f1Ndffz1sL7oHuWh975pSkJuNrleMrHZqV/vULsQJQhbem5msnjRp2NkQugUh6RM6oeMVI0/saJ/ahXiwU3eI4UmlBS+G3Q/cH0N9UJOSkoR5k30PT5eg9VVTCnKz0aV9Wni3U9n+tQtx2gBFPKtXr1ZtQSnJyckhB7iF9wd13rx5Qj3qPupbhL6/mlKQm40O7dNiqHYq0792IU57X4tn27Ztqi0oZfTo0agPI8AtPD+oL+7dK9ilvHt4nvpO7ijLy8ul6hPOxunt08JXO5XlX7sQ7+3tVW3BOCJ9AZ1Tp05hTV5eWAFuYX1QPzh7lu7hCdZva2uTqk+IxWntxy59f+1Uhn/tQpwgRJCcfB25uRfc3+elpoYd4BayF3zQoSMzWZ8IDF3fX930tQvxrKws1RaMY+3ataot2Epy8nWUlm7Go49uHhTkIll6//1adQQ66N9zzz1S9QmxOK392KUfSDsV6V+7EHf52cOZCI2KigrVFmzDCvCcnDo0NaWjsTEDgPhLf1VVVY7uaHTUr6qqkqpPiMVp7ccu/UDbqSj/2oV4h4Q9hSOdkydPqrZgC54BXlubjbKyUrhc/SsAiu4ILl68CMC5HY2O+lZNZekT4nFS+7FLP5h2KsK/diFOEKHgL8CByL30R/qEbHR/f52ur12I08Ik4lmwYIFqC1JhrA+PPLLVZ4AD4j+o06ZNk6rvTSToe9dUtD4hDye0H7v0Q2mn4fjXLsT9rWVNhEZPT49qC1LhPAqvv34PLl8ePWSAW4jsCIaaCumkjkZH/XCnl1KQq0V1+7FLP9R2Gqr/mJCOppDW1lbVFozjjTfewJQpU1TbkAAH0L/2+UcfTcHJk/ngfKi/W+PBli611VmojEhIx3cfftjn854dway8PGQJvnIVqP7Pd+7B9yJ8ESHiRpzSPmXqV1ZWIjc3V5q+N9qdiRNEICQnu7B+/R9w001n3Y8NHeAA0In+wHf+1/WOpmF/dyeckfT7VF+vwL4IO3FC+zRJX7sQT0xMVG3BOAoLC1VbEEpysgulpWXIzb2I++57BZHYUTutoyEIT3Rvn/70x48fL1XfGwpxAkVFRaotCMMKcGsQ23PPrYF1ST3S0L2jJMxG9/bpS/+mm26Squ+NdiFOG6CI5/nnn1dtQQjeAe5vEFukoHtHSZiN7u1zKP0333xTir4vtAtxghgKCnDf6N5REmaje/u0S98X2oV4TIx2A+odjwl7tI8adRWZmQ0U4D7QvSMjzEb39ump3ycho/yNUtcuxDMyMlRbMI6VK1eqthA2p09PwrPPPkwB7gc7OzKCCBZTgjw6M9PWP2S1C3G6Jy6eLVu2qLYQEsnJLnziE+fd358+PYkCfBhUX/ojCH+YEOTtV67YekVKuxAPd9Um4kb8bWLvVKx74J///DODgpwYHjs6MoIIFTuDvL2zU7h+QlSUrbeWtAtxgvAcxNbUlI6GBv3v6dsN3cMmnIxdQX61sVHLM35PtAvxkSNHqrZgHOvWrVNtIWBoFLo4KMgJJ2NHkI/KyBCub20oZdfnS7sQp7XTxVNeXq7aQkBQgIuHgpxwMrLbZ2J8vHD948ePu/9vx+dLuxDvlHAPI9I5ffq0agvDwhjH2rX+txMlQoOCnHAyug12u3LlilR9b7QLcSIy4ZzhwIF7cenSmIgPcB06MoIQiW5Bbqe+diGelpam2oJxLFq0SLUFnzD28eYlVVX5ePrpL0V0gAPQsqMhiHDRJWhnzJghVd8b7UKcppiJx6njDPq3E92IiRM/vtzPeWRuZuKJDh0ZQejYPkXot7e3S9X3RrsQv379umoLxvH222+rtnAD1iC2ceMu4VOf2j/ojDzS0aEjIwhd22e4+idOnJCq7412IU6Yj/co9GeffZjOwL1wekdGEDq3T530tQvxpKQk1RaMY9q0aaotuKFpZIGjU0dDRB66t89Q9SdOnChV3xvtQjw+Pl61BePIc8ha1xTgwePUjowgAP3bZyj6Y8aMkarvjXYh3tjYqNqCcezYsUO1BQBAdnYtbScaAk7syAjCQvf2Gax+sGOMwvWvXYgT5lJTMxFbtqylAA8Bp3VkBOGJ7u3TyfrahXhsbKxqC8aRk5Oj7NjJyS7cdNNZ9/c1NRMpwEPEyR0NQejePgPVT09Pl6rvjXYhHmqBCN8sX75cyXGte+CPPLJlUJAToeOUjowghkL39hmIfklJiVR9b7QL8fr6etUWjGPTpk22H9NzEFtjYwauXaPd6UThhI6MIHyhe/scTv/V/ful6nujXYj39fWptmAcXV1dth6PRqHLR3VHRhD+0L19+tPv6e6Wqu+NdiFO6A0FuH3o3lESZqN7+3SKvnYhnp2drdqCcaxfv96W49B2ovbjlI6GIIZC9/Y5lP7ChQul6PtCuxBvoY5COPv27bPlOJwz7N//SdpO1GZ07ygJs9G9fXrrHz16VIq+L7QL8c7OTtUWjOPcuXNS9T03Lzl1Ko+2E1WA7h0lYTa6t09P/VPnz0vR94V2IU7oRXKyC1/60tO4+eZT7sdoMxM12NmREUSwmBLkVbW1tv4hq12I0zxx8SxevFiKrjWIbezYy1iwgLYTdQJ2dWQEEQomBPnnFi2y9YqUdiHeLWD4PjGYuro64ZrJycmDRqFv3bqWzsAdgh0dGUGEip1B3i7h9mx0b6+tt5a0C3GXy6XagnEcOnRIqJ7L5UJpaSmNQncwdA+bcDJ2BfnVxkbh+idPnrT186VdiBPOxuVyoaysDDk5ORTgDoeCnHAydgT5qIwMbS/dW2gX4snJkR0Io0dPAGNM6NeXv/xlYVqFhYW4dOkSamtrKcA1gIKccDKy22difLxw/TyPMSF2fL60C/FI38Xs6tWzALhjv86ercGWLetRVlZGAa4JFOSEk9FtsFvWyMH7QMj2r12INzU1qbZAeJGc7MKECTXu78+enUBjFySiQ0dGECLRKcjfPXhQqr432oU44Sw8txP1DHJCHjp0ZAQhGp2C3E597UI8Li5OtQViAM/NTBoaMlFXR+va24GOHQ0ReejYPkXoZ2ZlSdX3RrsQT0tLU22BAO1GphIdOjKC0LV9hqs/u7hYqr432oX4tWvXVFuIeCjA1eP0jowgdG6f4egHsqGUSP/ahTjntHSnShjjePjhZynAHYCTOzKC0L19hqrf19cnVd8b7UKcMVq6UyWcM7z66gJcvDiWAtwBOLUjIwhA//YZin5UVOCxKsK/diE+0msOHmEPnpuXnDkzCRs3rqcAdwhO7MgIwkL39hms/sKFC6Xqe6NdiDc3N6u2EHFY24nm5X28xSRtZuIsnNaREYQnurfPYPQPVVRI1fdGuxDv6upSbSGi8NxO9JOffI22E3UwTurICMIb3dtnoPoN9fVS9b3RLsQJ+/AehU7biTofp3RkBDEUurdPJ+prF+Lp6emqLUQENI1MX5zY0RCEhe7tczj9O0pKpOp7o12Id3d3q7ZgPBTg+qO6IyMIf+jePv3p1wtYyyQY/9qFOG2sIZ+srHpkZDRSgGuO7h0lYTa6t09f+tXV1X5+Knx9b7QLcUI+586NxzPPfJ4C3AB07ygJs9G9fdqp7wupIc4Yu58x9hFjrJox9vgQz69ljH0w8PX2/9/evQdHdZ53HP8+3IQkQAghW1ggLgZqY2rHBFyKbQzFwZdpnKZ2azs24CQuk7bJtNNpk7bTpumkmTST/tHJNIkH09Rk4oAvdVx7fI2JMXYJCdgxGAUH8CUEFwQIWcKLuEi8/WN35eWglVarc86ed/X7zGiQVquHd149+/60Z8+e18yu6K9mdbVCJQrV1SlmzHi75+v9+5sU4GXC94VMypvv/RmsP3v27Ejq5xNZiJvZcODbwE3AHOBOM5sTuNs7wHXOucuBrwJr+qs7cuTIsIc65GVfA//Up354TpBL+fB9oZTy5nt/5tbvHj48kvr5RPlM/Cpgn3PubefcaWAD8IncOzjntjjn2jJfbgUm91f0/fffD32gQ1lwO9GWlgtLPSSJSBIO/YnkUy5B/vBzz8X6h2yUId4I/Cbn6wOZ2/L5LPBMhOORAJ2FPvTEtZCJFKMcgnzWBRfEekRqRIS1e7sqSK+X+zKzpaRD/Jo8318NrAaoq6tjzZr0UferrrqK+vp6nnrqKQCamppYvnw5a9euBWDUqFHcc889PP744xw+fBiAW2+9lX379rFjxw4AFi1axNixY3nuuecAmDFjBosXL+aBBx4AoKqqirvvvptHH32UY8eOAXD77beza9cumpubAbjuuusYMWIEGzduBGD27NksWLCABx98EIBx48Zxxx13sGHDBjoyv9i77rqLbdu2sWfPHgCWLVtGV1cXL730EgCXXXYZc+fO5aGHHgJgwoQJ3Hbbbflne4AU4P569tlnWbBgAe3t7T39M3PmTOomTuRnW7cCMKGujqsWLOD555/n7NmzDBs2jOXLl/Pzbds41tpKx8mTvPjBB0ytqaH14EEg3bc1NTVsy1w2sr6+niuvvLJna8URI0dy/bJlbN26teeI2KJFizh48CDvvPMOAJdeeimVlZWxzock0+bNm1m8eDGbN2/mxIkTACxZsoS9e/fy3nvvAXDFFVfQ3d3Nrl27gPQaPnXqVNr37+e+7du5csYMbr7+el588UVOnToFpNfK5uZmDh06BMC8efPo7Oxk9+7dAEyfPp1JkyaxZcsWIH1tkYULF/LCxo10Zd6i/NGFC1n/zDM0v/lmKI8nSL8/vPXoUUY4R/v+/Tx58CBXz53LW5lxDebx9Nprr+WdZ4tqa08z+13gK865GzJf/x2Ac+7rgftdDvwIuMk5t6e/uvPnz3fbt2+PYMR+SO/iNtjfmePee/+TyZPfizDAwxhnXPwaq3viiVAqtXZ0sH3fPubPnNnna27FsFtuwac59WOsvowTwujTKPszW/9La9fyjXvvDbV+NuTDHr/dcsurzrn5wdujPJy+DZhlZtPNbBRwB3DOb9XMmoDHgBWFBDjAkSNHQh/o0GO88ML1HDjQqGfgQ5hORpMki+PQ94W1taHXzz7TjuvxFVmIO+e6gM8DzwG7gYedc81m9jkz+1zmbl8G6oDvmNnrZjZ0n2LHwOzDzerffXcaa9d+VgE+xCnIJcmi7s/KigqvX4OHiN8n7px72jk32zl3sXPua5nb7nPO3Zf5/F7nXK1z7iOZj/MOFZw34AFsuC4fqq5OsXr1/cye/aucW7WZiSjIJdl8O9ltROBt0FGP37tErKurK/UQvJM9iW3SpEMsXbrpnGfk4h8fFjKRMPkU5NcvWxZp/SDvQlzvEx+Y4FnoP/jB3Tjn3a9dcviwkImEzZcg35o5oz2q+kHerebaxaxwehtZefJhIRPxsT/DqN/XE80oxu9diEthFODly4eFTMTX/vStvnchXltbW+oheKG29hjjx7+vAC9Tvi00MvT43J+Dqb9o0aJI6wd5F+LZq/ZI3w4cmML3v6/tRMtZkhcyEd/7s9j6BzNXQIyqfpB3IZ69fJ+cr7o6xcUXv9Xz9YEDUxTgZS6pC5kI+N+fxdTPXjI1qvpB3oW49C53O9HcIJfyl8SFTCTL9/5Men3vQnzMmDGlHkLi5J7E1tpax6FDDaUeksQs6QuNDG2+9+dA6l966aWR1g/yLsSHR7Dhus90FrpkJWkhEwnyvT8LrV/sLn7Fjt+7EG9vby/1EBJDAS5BSVnIRHrje38WUr+vbUPDqB/kXYhLluPOO9crwOU8SVjIRPLxvT+TVt+7EK+oqCj1EBLCeP75j2k7UelV0hYakVy+92df9RsaBn9O0kDG712Ijx07ttRDKCmzD3ce279/qrYTlbx8XyilvPnen/nqX3bZZZHWD/IuxI8ePVrqIZRMKpVi9erVXHLJmzm3ajtRyc/3hVLKm+/92Vv9jRs3RlI/H+9CfKhKpVKsW7eOSZMmcd11L2k7USmY7wullDff+zOu+vl4F+JD8S1m2QA/cuQIhw8f1naiMmC+L2RS3nzvz9z6J8+G/wSrbty4vN/zLgkmTJhQ6iHEKjfA6+vrWbdunV4Dl6LEuZCJDFS5BHllQ0Osf8h6F+JtbW2lHkJsggG+atUqUqlUqYclHiv1oT+RvpRDkHcfOxbrESnvQryrq6vUQ4jN+++/T3t7e0+AV1frGbgMXhwLmUix4gzyzgh2xRzW1RXrS0vehfhQ0tjYyMqVKxXgEjq9hi1JFleQt7S1efmMP5d3IV7ur4mnUin25bym2NjYqACXSCjIJcniCPILa2tDr3/ttdf21I/j8eVdiHd2dpZ6CJHJvga+fv36c4JcJCoKckmyqPuzsqIi9Pq//vWvez6P4/GlEE+I3JPY6urqmDRpUqmHJEOEglySzLeT3fbv3x9p/SDvQrwc9XYWug6hSz4+LGQiYfItyOOs712Il9u10xXgMlA+LjQig+VL0M6dOzfS+kHehXjuBiC+c86xYcMGBbgMiA8LmYiP/RlG/b6uKhrF+L0L8Y4yWljMjOXLlzN58mQFuBTMh4VMxNf+HGz9HTt2RFo/yLsQLwdnc66tO2XKFD7zmc8owGVAkr6Qifjcnz7V9y7ER48eXeohDEoqlWLNmjU0Nzf33FZOLxFIfHxaaGTo8b0/i63f2NgYaf0g70Lc52es2ZPYWlpaePnll895Ri5SjKQuZCLgf38WU3/WrFmR1g/yLsRbW1tLPYSiBM9CX7FiBcOGeTf9kkBJXMhEsnzvz4HW37RpU6T1g5QiMdDbyCRqSVvIRHL53p9Jru9diPd1+n4SKcAlLkleaER8789C61dVVUVaP8i7EPdtA5T29nY6OjoU4BKLpCxkIr3xvT8Lqb948eJI6wd5F+LHjh0r9RAG5KKLLtJ2ohKrJCxkIvn43p/91d+8eXOk9YO8C/Hu7u5SD6FfqVSKPXv29Hx90UUXKcAlVqVeyET64nt/9lX/xIkTkdYPGjHo/60MvPDYY3S2tBT1s6dOn6a1vZ26mhoqRo3izNmz7G1r42R3NxfX1FBTUTGosQXrixQqdyGYP3MmdePGeVVfypvv/Rl3/Xy8C/G6urrQa3a2tPDxyZOL/vnWjg6279vHnGnTePLQIU52d1M/ejSfbGqieuTIQY8vW7+vX6RIb3xfyKS8+d6fvdVfsmRJJPXz8e5weiqVKvUQzlM3bhxzpk1j/bvvcuTkSepHj2bVrFmhBHi2fn+/SJF8fD90KeXN9/4M1t+7d28k9fPxLsRPnjxZ6iGcJ3XmDE8eOkQnUAl8vKEhtADP6u8XKdIX3xdKKW++92du/d1vvRVJ/Xy8C/Gkcc7x0Ntv9zwDv3PaNH757ruRNYpIseJcyEQGqlyCfO/hw7H+IetdiI9LWJCZGTdMnsyU6mpWzZrFlAkT9IxEEiuuhUykGOUQ5H+wdGms6793Ie6cK/UQADibM47G6mo+PXt2zyF0HVqUJItjIRMpVpxB3nnqVOj1ayorY13/vQvx48ePh17z1OnTA7p/6swZ1rz5JrtyLjwT3E5UQS6RU+pKAAAL8klEQVRJpv6UJIsryFva2kKvv2vXrlgfX96FeBRa29sLnujUmTOs27uXls5OXmlpOecZeZAWSkky9ackWRxBfmFtrbeH7rO8C/HKysrQa9bV1BQ00dkAz57EtmLmTIYFnoGfV1sLpSSY+lOSLOr+rKyoCL1+U1NTz+dxPL4U4kDFqFH9TnQwwAfyPnAtlJJk6k9JMt9Odps6dWqk9YO8C/GoNkDpa6IHE+CF1BcZCB8WMpEw+RTkL7/8cqT1g7wL8Sjlm+iOM2foOHNm0Fdi00IpYfBhIRMJm09BHmd970J8xIhoL/fe20RPqqpi1axZoVxKVQulDJaPC40MPT72Zxj1x44dG2n9IO9CvLa2NvL/I3st9OcDQR7FtdC1UMpA+bCQifjan4Otf/XVV0daP8i7EI/qNfFc2Wuh74VzgjxMWihlMJK+kIn43J+Dqf/iiy9GWj/IuxDv7u6OtH7uSWwTR4/mmmnTEtkoIkleyER8789i658q8CpwYY3fuxCPUm9noUd9LXQtlDIYSV3IRMD//vShvnchPnHixEjq9vU2Mh9+kTJ0qT8lyXzvz4HWX7ZsWaT1g7wL8Siune6c4+Gc7UR7Ows9aY0ikkv9KUnme38OpH5zc3Ok9YO8C/FCX28YCDPjxilTaMpsJ5rvLPQkNYpIkPpTksz3/iy0/qFDhyKtH+RdiIcp9yS5SVVV3JOznWg+SWkUkd6oPyXJfO/PJNb3LsRrampCqZNKpbj//vvZuXNnz23B7UTzSeIvUiRL/SlJ5nt/9ld/3rx5kdYP8i7Ew3iLWSqVYt26dbS0tLBlyxZcH9uJ5lPqRhHpi/pTksz3/uyrfmdnZ6T1g7wL8Q8++GBQP58N8CNHjlBfX8+KFSsKfgYe5HsjSnlTf0qS+d6f+erv3r070vpB3oX4YAQDfNWqVVRXVw+qpu+NKOVN/SlJ5nt/xlk/H+9CvKqqqqifiyLAs5LwixTJx/eFTMqb7/0ZrD99+vRI6ufjXYhXVFQU9XPHjx/n+PHjoQd4VlyNIlIM3xdKKW++92du/YoxYyKpn493Id7W1lbUzzU0NLBq1apIAjwrjkYRKZaOGEmSlUuQP/jUU7H+IetdiA9EKpU65ySDhoaGyAI8S89IJMl0xEiSrByCfNYFF8S6/nsX4iML3NM7+xr4ww8/HNrZgoVSkEuS6YiRJFmcQd4ZwRVAmxoaYl3/vQvx8ePH93uf4ElsTU1Nfd7/1OnTYQ2vh4Jckkz9KUkWV5C3tLWFXn/hwoWxPr68C/HW1tY+v1/MWeit7e1eHroRGQz1pyRZHEF+YW1t6PVf2Lixp34cjy/vQvzs2bN5v1fs28jqamq8fQ1GZDDUn5JkUfdnZUVF6PW7zpzp+TyOx5d3IZ6Pc45HHnmkqPeBV4wa5fXJFCKDof6UJCuHk92irO9diNfX1/d6u5lx4403MnXq1KLeRub7L1KGDvWnDDU+rc/Lly+PtH6QdyHeEZiA3A1Rsu8FL/ZtZD41igxd6k8ZinxZn3/xi19EWj/IuxA/lfOWgFQqxZo1a86ZtGI3M8nypVFk6FJ/ig987M8w6h85ciTS+kHehXhW9iS2w4cPs3Xr1lC2KM3yoVFk6FJ/ig987U/f6kca4mZ2o5n9ysz2mdnf9vJ9M7NvZb6/08z63U19/Pjx552FvnLlSoYPHx7q2H37RcrQov6UpPO5PwdTf8GCBZHWD4osxM1sOPBt4CZgDnCnmc0J3O0mYFbmYzXw3f7qnj59OrLdyIKS3Cgi6k9JMt/7s9j67e3tkdYPivKZ+FXAPufc286508AG4BOB+3wC+L5L2wqMN7NJfRXt6OiIJcCzktooIqD+lGTzvT+Lqb9nz55I6wdFGeKNwG9yvj6QuW2g9zlPXAGelcRGEclSf0qS+d6fSa9vzrnQBwVgZn8E3OCcuzfz9QrgKufcF3Lu8xTwdefcK5mvNwJfdM69Gqi1mvThdoC5wK5IBj10TQSOlnoQZUZzGj7Nafg0p+GLak6nOufOu1DKiAj+o6wDwJScrycD/1fEfXDOrQHWAJjZdufc/HCHOrRpTsOnOQ2f5jR8mtPwxT2nUR5O3wbMMrPpZjYKuAN4InCfJ4CVmbPUFwLtzrmDEY5JRESkbET2TNw512VmnweeA4YD33PONZvZ5zLfvw94GrgZ2AecAD4d1XhERETKTZSH03HOPU06qHNvuy/ncwf8+QDLrglhaHIuzWn4NKfh05yGT3MavljnNLIT20RERCRa3l52VUREZKhLbIhHccnWoa6AOb0rM5c7zWyLmV1RinH6pL85zbnfAjPrNrPb4hyfjwqZUzNbYmavm1mzmb0U9xh9U8Bjv8bMnjSzHZk51flJfTCz75nZYTPr9e3OseaTcy5xH6RPhHsLmAGMAnYAcwL3uRl4BjBgIfCzUo87yR8FzukioDbz+U2a08HPac79fkL6/JDbSj3uJH8U2KfjgV8CTZmvLyj1uJP8UeCc/j3wjczn9cAxYFSpx57UD2AxMA/Ylef7seVTUp+JR3LJ1iGu3zl1zm1xzrVlvtxK+n37kl8hfQrwBeC/gcNxDs5Thczpp4DHnHP7AZxzmte+FTKnDhhr6b2cx5AO8a54h+kP59xm0nOUT2z5lNQQj+ySrUPYQOfrs6T/kpT8+p1TM2sEPgnchxSikD6dDdSa2SYze9XMVsY2Oj8VMqf/AVxK+mJbbwB/4Zw7G8/wylJs+RTpW8wGwXq5LXgafSH3kQ8VPF9mtpR0iF8T6Yj8V8ic/jvwJedcd/pJjvSjkDkdAXwUWAZUAj81s63OucJ3nhhaCpnTG4DXgd8DLgZ+bGYvO+d0sfzixJZPSQ3x0C7ZKj0Kmi8zuxxYC9zknGuNaWy+KmRO5wMbMgE+EbjZzLqcc4/HM0TvFPrYP+qcSwEpM9sMXAEoxHtXyJx+GvhXl35Bd5+ZvQNcAvw8niGWndjyKamH03XJ1vD1O6dm1gQ8BqzQs5qC9DunzrnpzrlpzrlpwKPAnynA+1TIY/9/gGvNbISZVQG/A+yOeZw+KWRO95M+soGZXQj8FvB2rKMsL7HlUyKfiTtdsjV0Bc7pl4E64DuZZ45dTpsj5FXgnMoAFDKnzrndZvYssBM4C6x1zmlnwzwK7NOvAg+Y2RukDwV/yTmn3c3yMLP1wBJgopkdAP4JGAnx55Ou2CYiIuKppB5OFxERkX4oxEVERDylEBcREfGUQlxERMRTCnERERFPKcRFPGdmDWa2wczeMrNfmtnTZja7iDrXZnawet3MGs3s0Tz322RmeuuhSAIoxEU8ltmw4kfAJufcxc65OaR3pLqwiHJ3Af/mnPuIc+4955y2TRVJOIW4iN+WAmdyLyzjnHsdeMXMvmlmu8zsDTO7HXr24d5kZo+a2Ztm9mDmqlL3An8MfDlz27TsXslmVpl5pr/TzB4ifb1yMt9bbmY/NbPXzOwRMxuTuf1dM/vnzO1vmNklmdvHmNl/ZW7baWa39lVHRPqmEBfx21zg1V5u/0PgI6SvKX498M2crRCvBP4SmEN6j+mrnXNrSV8q8m+cc3cFav0pcMI5dznwNdKbj2BmE4F/AK53zs0DtgN/lfNzRzO3fxf468xt/0j6EpS/nan3kwLqiEgeibzsqogM2jXAeudcN9BiZi8BC4AO4OfOuQMAZvY6MA14pY9ai4FvATjndprZzsztC0n/IfC/mcv0jgJ+mvNzj2X+fZX0HxWQ/oPijuwdnHNtZvb7/dQRkTwU4iJ+awZ6e+26r31PT+V83k1h60Bv12c24MfOuTv7+X9y/w/rpVZ/dUQkDx1OF/HbT4AKM/uT7A1mtgBoA243s+FmVk/62XSx20puJn3SG2Y2F7g8c/tW4Gozm5n5XlUBZ8U/D3w+Z6y1RdYRERTiIl7L7P/8SeBjmbeYNQNfAX5IepevHaSD/ovOuUNF/jffBcZkDqN/kcwfA865I8A9wPrM97aS3oO6L/8C1GZOuNsBLC2yjoigXcxERES8pWfiIiIinlKIi4iIeEohLiIi4imFuIiIiKcU4iIiIp5SiIuIiHhKIS4iIuIphbiIiIin/h/1H5xmyj2OWAAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "if radio_input7.value=='No CV':\n",
    "    cv=0\n",
    "elif radio_input7.value=='5-fold CV':\n",
    "    cv=int(5)\n",
    "elif radio_input7.value=='10-fold CV':\n",
    "    cv=int(10)\n",
    "    \n",
    "if data_icu:\n",
    "    model=dl_train.DL_models(data_icu,diag_flag,proc_flag,out_flag,chart_flag,med_flag,False,radio_input6.value,cv,oversampling=radio_input8.value=='True',model_name='attn_icu_read',train=True)\n",
    "else:\n",
    "    model=dl_train.DL_models(data_icu,diag_flag,proc_flag,False,False,med_flag,lab_flag,radio_input6.value,cv,oversampling=radio_input8.value=='True',model_name='attn_icu_read',train=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "driven-factor",
   "metadata": {},
   "source": [
    "## 10. Running BEHRT\n",
    "Below we integrate the implementation of BEHRT in our pipeline.\n",
    "We perform pre-procesing needed to run BEHRT model. https://github.com/deepmedicine/BEHRT\n",
    "\n",
    "Few things to note before running BEHRT -\n",
    "- The numerical values are binned into quantiles.\n",
    "- BEHRT has recommended maximum number of events per sample to be 512. \n",
    "    So feature selection is important so that number of events per sample does not exceed 512.\n",
    "- The model is quite computationally heavy so it requires a GPU.\n",
    "\n",
    "The output files for BEHRT will be saved in ./data/behrt/ folder\n",
    "\n",
    "**Please run below cell to to pre-process and run BEHRT on the selected cohort**"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "aggressive-break",
   "metadata": {},
   "outputs": [],
   "source": [
    "if data_icu:\n",
    "    token=tokenization.BEHRT_models(data_icu,diag_flag,proc_flag,out_flag,chart_flag,med_flag,False)\n",
    "    tokenized_src, tokenized_age, tokenized_gender, tokenized_ethni, tokenized_ins, tokenized_labels=token.tokenize()\n",
    "else:\n",
    "    token=tokenization.BEHRT_models(data_icu,diag_flag,proc_flag,False,False,med_flag,lab_flag)\n",
    "    tokenized_src, tokenized_age, tokenized_gender, tokenized_ethni, tokenized_ins, tokenized_labels=token.tokenize()\n",
    "    \n",
    "behrt_train.train_behrt(tokenized_src, tokenized_age, tokenized_gender, tokenized_ethni, tokenized_ins, tokenized_labels)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "reserved-objective",
   "metadata": {},
   "source": [
    "### EVALUATION AS STANDALONE MODULE\n",
    "Below cell shows an exaple of how evaluation module can be used as a standalone module.\n",
    "\n",
    "evaluation.Loss class can be instantiated and model output and ground truth can be passed to it to obtain results.\n",
    "\n",
    "In the example below we captured model output and ground truth in a file and used that file to read the data.\n",
    "\n",
    "In function definition ***loss(prob,truth,logits,False)***\n",
    "\n",
    "prob -> List of Output predicted probabilities of case being positive\n",
    "\n",
    "truth -> List of ground truth labels\n",
    "\n",
    "logits -> List of logits obtained from last fully connected layer before applying softmax.sigmoid function in the model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "streaming-integration",
   "metadata": {},
   "outputs": [],
   "source": [
    "if torch.cuda.is_available():\n",
    "    device='cuda:0'\n",
    "#device='cpu'\n",
    "loss=evaluation.Loss(device,acc=True,ppv=True,sensi=True,tnr=True,npv=True,auroc=True,aurocPlot=True,auprc=True,auprcPlot=True,callb=True,callbPlot=True)\n",
    "with open(\"./data/output/outputDict\", 'rb') as fp:\n",
    "    outputDict=pickle.load(fp)\n",
    "prob=list(outputDict['Prob'])\n",
    "truth=list(outputDict['Labels'])\n",
    "logits=list(outputDict['Logits'])\n",
    "#print(torch.tensor(prob))\n",
    "print(\"======= TESTING ========\")\n",
    "loss(prob,truth,logits,train=False,standalone=True)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "designing-works",
   "metadata": {},
   "source": [
    "### 11. FAIRNESS EVALUATION\n",
    "In train and testing step we save output files in **./data/output/** folder.\n",
    "\n",
    "This file conatins list of demographic variables included in training and testing of the model.\n",
    "\n",
    "It also contains the ground truth labels and predicted probability for each sample.\n",
    "\n",
    "We use the above saved data to perform fairness evaluation of the results obtained from model testing.\n",
    "\n",
    "This module can be used as stand-alone module also.\n",
    "\n",
    "Please create a file that contains predicted probabilites form the last sigmoid layer in column named **Prob** and\n",
    "ground truth labels for each sample in column named **Labels**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "civilian-direction",
   "metadata": {},
   "outputs": [],
   "source": [
    "fairness.fairness_evaluation(inputFile='outputDict',outputFile='fairnessReport')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "efficient-distinction",
   "metadata": {},
   "source": [
    "### 12. MODEL CALLIBRATION\n",
    "\n",
    "Please run below cell if you want to callibrate predicted probabilites of the model on test data.\n",
    "It will use the output saved during the testing of the model.\n",
    "\n",
    "The file is saved in **./data/output/**.\n",
    "\n",
    "This module can be used as stand-alone module also.\n",
    "\n",
    "Please create a file that contain predicted logits form the last fully connected layer in column named **Logits** and <br>ground truth labels for each sample in a column named **Labels**."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "secure-flavor",
   "metadata": {},
   "outputs": [],
   "source": [
    "callibrate_output.callibrate(inputFile='outputDict',outputFile='callibratedResults')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "fleet-canyon",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "hide_input": false,
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.19"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}