qrmb_medical - autoencoder - diagnosis.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Restricted Boltzmann Machine using quantum annealing on D-wave's QPU"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Imports"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/lib/python3/dist-packages/scipy/__init__.py:146: UserWarning: A NumPy version >=1.17.3 and <1.25.0 is required for this version of SciPy (detected version 1.25.2\n",
      "  warnings.warn(f\"A NumPy version >={np_minversion} and <{np_maxversion}\"\n"
     ]
    }
   ],
   "source": [
    "import numpy as np\n",
    "import timeit\n",
    "import matplotlib.pyplot as plt\n",
    "from tqdm import tqdm_notebook as tqdm\n",
    "\n",
    "from qrbm.EncodedQRBM import QRBM\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "%matplotlib inline\n",
    "plt.rcParams['image.cmap'] = 'gray'\n",
    "\n",
    "import pandas as pd\n",
    "\n",
    "from skimage import data, color\n",
    "from skimage.transform import rescale, resize, downscale_local_mean\n",
    "from skimage import img_as_bool\n",
    "\n",
    "import cv2 as cv\n",
    "\n",
    "import random"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Presets\n",
    "\n",
    "Here you can set hyperparameters for learning:\n",
    "\n",
    " - digits: list of digits to train on\n",
    " - THRESHOLD: how bright parts of image become 1 (darker become 0)\n",
    " - n_hidden: number of neurons in hidden layer\n",
    " - qpu: whether to use real D-wave's qpu (requires [setup](https://docs.ocean.dwavesys.com/en/stable/overview/sapi.html)) or a local simulator\n",
    " - epochs: number of epochs\n",
    " - lr: learning rate\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "n_hidden = 50\n",
    "qpu = False\n",
    "epochs = 50\n",
    "lr = 0.1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## reading dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "data = pd.read_csv('encoded_data.csv')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## prepare train and test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "cols = data.columns"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "cols = [x for x in cols if not x.startswith('Disease')]"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "cols = [x for x in cols if not '?' in x]"
   ]
  },
  {
   "cell_type": "raw",
   "metadata": {},
   "source": [
    "cols = [x for x in cols if not 'Treatment' in x]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Index(['Sex', 'Stoke Risk (ABCD2 > 4)', 'Dizziness Type',\n",
       "       'Triggered by turning in bed?', 'SG Vertical N-None',\n",
       "       'SG Vertical N-Down-beating_vs_Up-beating', 'SG Horizintal N-None',\n",
       "       'SG Horizintal N-Left-beating_vs_Right-beating', 'SG Torsional N-None',\n",
       "       'SG Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Right Gaze Vertical N', 'Right G Horizintal N',\n",
       "       'Right G Torsional N-None',\n",
       "       'Right G Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Left Gaze \"Induced\" Vertical Nystagmus',\n",
       "       'Left Gaze \"Induced\" Horizintal Nystagmus',\n",
       "       'Left Gaze \"Induced\" Torsional Nystagmus-None',\n",
       "       'Left Gaze \"Induced\" Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Up Gaze \"Induced\" Vertical Nystagmus',\n",
       "       'Up Gaze \"Induced\" Horizintal Nystagmus-None',\n",
       "       'Up Gaze \"Induced\" Horizintal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Up Gaze \"Induced\" Torsional Nystagmus-None',\n",
       "       'Up Gaze \"Induced\" Torsional Nystagmus-Top pole to right ear torsional nystamgus_vs_Top-pole-to-L-ear torsional',\n",
       "       'Down Gaze \"Induced\" Vertical Nystagmus',\n",
       "       'Down Gaze \"Induced\" Horizintal Nystagmus-None',\n",
       "       'Down Gaze \"Induced\" Horizintal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Down Gaze \"Induced\" Torsional Nystagmus',\n",
       "       'Right Dix-Hallpike Induced Vertical Nystagmus-None',\n",
       "       'Right Dix-Hallpike Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Right Dix-Hallpike Induced Horizontal Nystagmus-None',\n",
       "       'Right Dix-Hallpike Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Right Dix-Hallpike Induced Torsional Nystagmus-None',\n",
       "       'Right Dix-Hallpike Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Left Dix-Hallpike Induced Vertical Nystagmus-None',\n",
       "       'Left Dix-Hallpike Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Left Dix-Hallpike Induced Horizontal Nystagmus-None',\n",
       "       'Left Dix-Hallpike Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Left Dix-Hallpike Induced Torsional Nystagmus-None',\n",
       "       'Left Dix-Hallpike Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Right Roll Induced Vertical Nystagmus-None',\n",
       "       'Right Roll Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Right Roll Induced Horizontal Nystagmus-None',\n",
       "       'Right Roll Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Right Roll Induced Torsional Nystagmus-None',\n",
       "       'Right Roll Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Left Roll Induced Vertical Nystagmus-None',\n",
       "       'Left Roll Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Left Roll Induced Horizontal Nystagmus-None',\n",
       "       'Left Roll Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Left Roll Induced Torsional Nystagmus-None',\n",
       "       'Left Roll Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional',\n",
       "       'Bow Induced Vertical Nystagmus-None',\n",
       "       'Bow Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Bow Induced Horizontal Nystagmus-None',\n",
       "       'Bow Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Bow Induced Torsional Nystagmus-None',\n",
       "       'Bow Induced Torsional Nystagmus-Torsional Top-Pole-to-the-Left-Ear_vs_Torsional Top-Pole-to-the-Right-Ear',\n",
       "       'Lean Induced Vertical Nystagmus-None',\n",
       "       'Lean Induced Vertical Nystagmus-Down-beating_vs_Up-beating',\n",
       "       'Lean Induced Horizontal Nystagmus-None',\n",
       "       'Lean Induced Horizontal Nystagmus-Left-beating_vs_Right-beating',\n",
       "       'Lean Induced Torsional Nystagmus-None',\n",
       "       'Lean Induced Torsional Nystagmus-Torsional Top-Pole-to-the-Left-Ear_vs_Torsional Top-Pole-to-the-Right-Ear',\n",
       "       'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused vertical-No eye motion',\n",
       "       'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused vertical-Left eye moves down (L hyper)_vs_Left eye moves up (R hyper)',\n",
       "       'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-No eye motion',\n",
       "       'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-L eye moves up (L hyper)_vs_R eye moves down (R hyper)',\n",
       "       'Head impulse test to the left showed',\n",
       "       'Head impulse test to the right showed', 'Gait',\n",
       "       'Finger-nose-finger left hand-Abnormal',\n",
       "       'Finger-nose-finger left hand-None_vs_Normal',\n",
       "       'Finger-nose-finger right hand-Abnormal',\n",
       "       'Finger-nose-finger right hand-None_vs_Normal',\n",
       "       'Hearing loss right ear', 'Hearing loss left ear'],\n",
       "      dtype='object')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "cols"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data[cols]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Sex</th>\n",
       "      <th>Stoke Risk (ABCD2 &gt; 4)</th>\n",
       "      <th>Dizziness Type</th>\n",
       "      <th>Triggered by turning in bed?</th>\n",
       "      <th>SG Vertical N-None</th>\n",
       "      <th>SG Vertical N-Down-beating_vs_Up-beating</th>\n",
       "      <th>SG Horizintal N-None</th>\n",
       "      <th>SG Horizintal N-Left-beating_vs_Right-beating</th>\n",
       "      <th>SG Torsional N-None</th>\n",
       "      <th>SG Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional</th>\n",
       "      <th>...</th>\n",
       "      <th>Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-L eye moves up (L hyper)_vs_R eye moves down (R hyper)</th>\n",
       "      <th>Head impulse test to the left showed</th>\n",
       "      <th>Head impulse test to the right showed</th>\n",
       "      <th>Gait</th>\n",
       "      <th>Finger-nose-finger left hand-Abnormal</th>\n",
       "      <th>Finger-nose-finger left hand-None_vs_Normal</th>\n",
       "      <th>Finger-nose-finger right hand-Abnormal</th>\n",
       "      <th>Finger-nose-finger right hand-None_vs_Normal</th>\n",
       "      <th>Hearing loss right ear</th>\n",
       "      <th>Hearing loss left ear</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>...</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>-1</td>\n",
       "      <td>1</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>10 rows × 76 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   Sex  Stoke Risk (ABCD2 > 4)  Dizziness Type  Triggered by turning in bed?  \\\n",
       "0    0                       1               0                             0   \n",
       "1    1                       1               0                             0   \n",
       "2    1                       1               0                             0   \n",
       "3    1                       0               0                             0   \n",
       "4    0                       0               0                             0   \n",
       "5    1                       1               0                             0   \n",
       "6    1                       1               0                             0   \n",
       "7    0                       1               0                             0   \n",
       "8    1                       0               0                             0   \n",
       "9    1                       1               0                             0   \n",
       "\n",
       "   SG Vertical N-None  SG Vertical N-Down-beating_vs_Up-beating  \\\n",
       "0                   0                                         0   \n",
       "1                   0                                         0   \n",
       "2                   0                                         0   \n",
       "3                   0                                         0   \n",
       "4                   0                                         0   \n",
       "5                   0                                         0   \n",
       "6                   0                                         0   \n",
       "7                   0                                         0   \n",
       "8                   0                                         0   \n",
       "9                   0                                         0   \n",
       "\n",
       "   SG Horizintal N-None  SG Horizintal N-Left-beating_vs_Right-beating  \\\n",
       "0                     0                                              0   \n",
       "1                     0                                              0   \n",
       "2                     0                                              0   \n",
       "3                     0                                              0   \n",
       "4                     0                                              0   \n",
       "5                     0                                              0   \n",
       "6                     0                                              0   \n",
       "7                     0                                              0   \n",
       "8                     0                                              0   \n",
       "9                     0                                              0   \n",
       "\n",
       "   SG Torsional N-None  \\\n",
       "0                    0   \n",
       "1                    0   \n",
       "2                    0   \n",
       "3                    0   \n",
       "4                    0   \n",
       "5                    0   \n",
       "6                    0   \n",
       "7                    0   \n",
       "8                    0   \n",
       "9                    0   \n",
       "\n",
       "   SG Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional  \\\n",
       "0                                                  0                           \n",
       "1                                                  0                           \n",
       "2                                                  0                           \n",
       "3                                                  0                           \n",
       "4                                                  0                           \n",
       "5                                                  0                           \n",
       "6                                                  0                           \n",
       "7                                                  0                           \n",
       "8                                                  0                           \n",
       "9                                                  0                           \n",
       "\n",
       "   ...  \\\n",
       "0  ...   \n",
       "1  ...   \n",
       "2  ...   \n",
       "3  ...   \n",
       "4  ...   \n",
       "5  ...   \n",
       "6  ...   \n",
       "7  ...   \n",
       "8  ...   \n",
       "9  ...   \n",
       "\n",
       "   Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-L eye moves up (L hyper)_vs_R eye moves down (R hyper)  \\\n",
       "0                                                  0                                                                                                \n",
       "1                                                  0                                                                                                \n",
       "2                                                  0                                                                                                \n",
       "3                                                  0                                                                                                \n",
       "4                                                  0                                                                                                \n",
       "5                                                  0                                                                                                \n",
       "6                                                  0                                                                                                \n",
       "7                                                  0                                                                                                \n",
       "8                                                  0                                                                                                \n",
       "9                                                  0                                                                                                \n",
       "\n",
       "   Head impulse test to the left showed  \\\n",
       "0                                     0   \n",
       "1                                     0   \n",
       "2                                     0   \n",
       "3                                     0   \n",
       "4                                     0   \n",
       "5                                     0   \n",
       "6                                     0   \n",
       "7                                     0   \n",
       "8                                     0   \n",
       "9                                     0   \n",
       "\n",
       "   Head impulse test to the right showed  Gait  \\\n",
       "0                                      0     0   \n",
       "1                                      0     0   \n",
       "2                                      0     0   \n",
       "3                                      0     0   \n",
       "4                                      0     0   \n",
       "5                                      0     0   \n",
       "6                                      0     0   \n",
       "7                                      0     0   \n",
       "8                                      0     0   \n",
       "9                                      0     0   \n",
       "\n",
       "   Finger-nose-finger left hand-Abnormal  \\\n",
       "0                                      1   \n",
       "1                                      1   \n",
       "2                                      1   \n",
       "3                                      1   \n",
       "4                                      1   \n",
       "5                                      1   \n",
       "6                                      1   \n",
       "7                                      1   \n",
       "8                                      1   \n",
       "9                                      1   \n",
       "\n",
       "   Finger-nose-finger left hand-None_vs_Normal  \\\n",
       "0                                           -1   \n",
       "1                                           -1   \n",
       "2                                           -1   \n",
       "3                                           -1   \n",
       "4                                           -1   \n",
       "5                                           -1   \n",
       "6                                           -1   \n",
       "7                                           -1   \n",
       "8                                           -1   \n",
       "9                                           -1   \n",
       "\n",
       "   Finger-nose-finger right hand-Abnormal  \\\n",
       "0                                       1   \n",
       "1                                       1   \n",
       "2                                       1   \n",
       "3                                       1   \n",
       "4                                       1   \n",
       "5                                       1   \n",
       "6                                       1   \n",
       "7                                       1   \n",
       "8                                       1   \n",
       "9                                       1   \n",
       "\n",
       "   Finger-nose-finger right hand-None_vs_Normal  Hearing loss right ear  \\\n",
       "0                                            -1                       1   \n",
       "1                                            -1                       1   \n",
       "2                                            -1                       1   \n",
       "3                                            -1                       1   \n",
       "4                                            -1                       1   \n",
       "5                                            -1                       1   \n",
       "6                                            -1                       1   \n",
       "7                                            -1                       1   \n",
       "8                                            -1                       1   \n",
       "9                                            -1                       1   \n",
       "\n",
       "   Hearing loss left ear  \n",
       "0                      1  \n",
       "1                      1  \n",
       "2                      1  \n",
       "3                      1  \n",
       "4                      1  \n",
       "5                      1  \n",
       "6                      1  \n",
       "7                      1  \n",
       "8                      1  \n",
       "9                      1  \n",
       "\n",
       "[10 rows x 76 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data[:10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(95, 76)"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "indices = list(range(len(data)))\n",
    "random.shuffle(indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "95"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(indices)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "85 10\n"
     ]
    }
   ],
   "source": [
    "num_train = int(len(data) * .9)\n",
    "train = indices[:num_train]\n",
    "test = indices[num_train:]\n",
    "print(len(train), len(test))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "X_train = data.values[train]\n",
    "X_test = data.values[test]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(85, 76)"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import tracemalloc \n",
    "  \n",
    "# tracemalloc.start() \n",
    "\n",
    "bm = QRBM(X_train[0], n_hidden=n_hidden, qpu=qpu)\n",
    "bm.tqdm = tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "762d432371a74215be5b2976e393f331",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/50 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "81\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "bm.train(X_train, epochs = epochs, lr = lr, lr_decay = 0.1)\n",
    "\n",
    "# snapshot = tracemalloc.take_snapshot() \n",
    "# top_stats = snapshot.statistics('lineno') \n",
    "  \n",
    "# for stat in top_stats[:10]: \n",
    "#    print(stat)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cs:2, lr:0.5\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7befe30bd50b40878b680eb8f5ece812",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/300 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "16\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cs:2, lr:0.1\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f23f22096ff8420c935b5bf8ab7e05e1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/300 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "56\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cs:2, lr:0.05\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6eade72c581149ba8c835585dfe052a2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/300 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "70\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cs:2, lr:0.01\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0acfa916f4f74da8ae9fdb20f2157131",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/300 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "50\n"
     ]
    },
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m/tmp/ipykernel_25913/4115735907.py\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      3\u001b[0m         \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34mf'cs:{cs}, lr:{lr}'\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m         \u001b[0mbm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m         \u001b[0mbm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mX_train\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mepochs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m300\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlr\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlr\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mepoch_drop\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m100\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m~/QRBM/qrbm/EncodedQRBM.py\u001b[0m in \u001b[0;36mtrain\u001b[0;34m(self, training_data, epochs, lr, lr_decay, epoch_drop, momentum, batch_size)\u001b[0m\n\u001b[1;32m    122\u001b[0m             \u001b[0;31m# 1.2 Compute the probabilites of the hidden units\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    123\u001b[0m             \u001b[0;31m# persisntent CD takes v from previous iterations\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 124\u001b[0;31m             h = self.sampler.sample_hidden(old_v,\n\u001b[0m\u001b[1;32m    125\u001b[0m                                            \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mhidden_bias\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    126\u001b[0m                                            \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mw\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/QRBM/qrbm/persistent_sampler.py\u001b[0m in \u001b[0;36msample_hidden\u001b[0;34m(self, v_states, h_biases, weights, chain_strength, num_reads, mask)\u001b[0m\n\u001b[1;32m     83\u001b[0m             \u001b[0mfeed_dict\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34mf'bh{j}'\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mh_biases\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mj\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     84\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 85\u001b[0;31m         \u001b[0mbqm\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvhH_model\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mto_bqm\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeed_dict\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     86\u001b[0m         \u001b[0msampleset\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msampler\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msample\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbqm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mchain_strength\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mchain_strength\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnum_reads\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mnum_reads\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     87\u001b[0m         \u001b[0msolution1\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msampleset\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mfirst\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msample\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m~/.local/lib/python3.10/site-packages/dimod/binary/binary_quadratic_model.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, offset, vartype, dtype, *args)\u001b[0m\n\u001b[1;32m    155\u001b[0m     \u001b[0;34m\"\"\"The default dtype used to construct the class.\"\"\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    156\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 157\u001b[0;31m     def __init__(self, *args,\n\u001b[0m\u001b[1;32m    158\u001b[0m                  \u001b[0moffset\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mOptional\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mBias\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    159\u001b[0m                  \u001b[0mvartype\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mOptional\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mVartypeLike\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "for cs in [2, 3, 4, 5, 6, 10]:\n",
    "    for lr in [0.5, 0.1, 0.05, 0.01, 0.005]:\n",
    "        print(f'cs:{cs}, lr:{lr}')\n",
    "        bm.cs = cs\n",
    "        bm.train(X_train, epochs = 300, lr = lr, epoch_drop=100)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bm.hidden_bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bm.visible_bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "bm.save('test_model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bm.load('test_model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Load raw dat"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data has 95 rows and 54 columns.\n",
      "Creating mask to determine which rows can be dropped\n",
      "Currently 0 rows are being dropped\n",
      "=====================\n",
      "Checking for entries that are droppable by being rare...\n",
      "0 rows are being dropped\n",
      "=======================\n",
      "Checking for unknown values that require warnings...\n",
      "=======================\n",
      "Checking for columns that lack signal...\n",
      "=======================\n",
      "Data has 95 rows and 54 columns.\n"
     ]
    }
   ],
   "source": [
    "from parsing_data import load_data, clean_data, load_config\n",
    "\n",
    "config_file = 'config.json'\n",
    "data_file = 'Simplified Dataset.xlsx'\n",
    "sheet = 'Copy of Cases-Use'\n",
    "\n",
    "config = load_config(config_file)\n",
    "raw_data = load_data(data_file, sheet)\n",
    "df = clean_data(raw_data, config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0     BPPV (PC) R\n",
       "1     BPPV (PC) L\n",
       "2     BPPV (PC) L\n",
       "3     BPPV (PC) R\n",
       "4     BPPV (PC) R\n",
       "         ...     \n",
       "90         Stroke\n",
       "91         Stroke\n",
       "92         Stroke\n",
       "93         Stroke\n",
       "94         Stroke\n",
       "Name: Disease, Length: 95, dtype: object"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['Disease']"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.preprocessing import OneHotEncoder, OrdinalEncoder, TargetEncoder\n",
    "\n",
    "encoder = OrdinalEncoder().fit(df[['Disease']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "y = encoder.transform(df[['Disease']])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "y_train = y.ravel()[train]\n",
    "y_test = y.ravel()[test]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([4., 1., 1., 2., 4., 5., 5., 4., 4., 4.])"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_test"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train a multi-class classifier"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Compile the training data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(10, 76)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_test.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([ 0,  0,  1,  1,  0,  0,  1,  0,  0,  0,  0,  0,  0,  0,  0,  1,  0,\n",
       "        0,  1,  1,  0,  0,  0,  0,  1,  0,  1,  0,  0,  1,  0,  0,  0,  0,\n",
       "        0,  1,  0,  0,  0,  0,  0,  1, -1,  0,  0,  0,  0,  1,  0,  0,  0,\n",
       "        0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  0,  1,  0,  1,  0,  1,\n",
       "        0,  0,  1, -1,  1, -1,  1,  1])"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X_train[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "50"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(bm.sample_hidden(X_train[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "a = bm.sample_hidden(X_train[0])\n",
    "b = bm.sample_hidden(X_train[0])\n",
    "a == b"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoded_train = np.array([bm.sample_hidden(x) for x in X_train])\n",
    "encoded_test = np.array([bm.sample_hidden(x) for x in X_test])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(85, 50)"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "encoded_train.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(85,)"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train.shape"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Train a model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.linear_model import LogisticRegression\n",
    "clf = LogisticRegression(random_state=0)\n",
    "\n",
    "clf = clf.fit(encoded_train, y_train)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([['Vestibular Neuritis R']], dtype=object)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "encoder.inverse_transform(np.array([[6.]]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00235359 0.00273424 0.01902502 0.01240163 0.95813286 0.00152153\n",
      "  0.00383112]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.25943161 0.57638271 0.02838741 0.03420375 0.04189771 0.03610004\n",
      "  0.02359678]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.39377541 0.30714618 0.02896548 0.01416426 0.21751309 0.02740455\n",
      "  0.01103102]]\n",
      "0.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.01663718 0.0143994  0.77663189 0.13665732 0.03683247 0.01128171\n",
      "  0.00756002]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[5.48801537e-04 9.04686890e-04 1.29192480e-03 5.09478370e-03\n",
      "  9.90899648e-01 1.62576751e-04 1.09757828e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.04085003 0.02778818 0.01278251 0.00696172 0.15826548 0.73428459\n",
      "  0.01906751]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.02589527 0.01768724 0.00529877 0.01123982 0.29097365 0.63744925\n",
      "  0.011456  ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[9.55870466e-04 2.19728587e-03 9.00489289e-03 5.99053312e-03\n",
      "  9.78395181e-01 1.22176421e-03 2.23447260e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[6.61975468e-03 1.97530576e-03 6.74434104e-04 1.24960129e-03\n",
      "  9.83103015e-01 1.89913821e-03 4.47875082e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01671844 0.00746667 0.00386532 0.00170363 0.87085366 0.09227314\n",
      "  0.00711915]]\n",
      "1.0\n",
      "===\n",
      "9 1 0.9\n"
     ]
    }
   ],
   "source": [
    "g = 0\n",
    "b = 0\n",
    "for x, y in zip(encoded_test, y_test.tolist()):\n",
    "    x = x.reshape(1, -1)\n",
    "    y = np.array([[y]])\n",
    "    print(encoder.inverse_transform(y))\n",
    "    pred_indx = clf.predict(x)\n",
    "    y_est = pred_indx.reshape((1, 1))\n",
    "    print(encoder.inverse_transform(y_est), pred_indx)\n",
    "    if y == y_est:\n",
    "        g += 1\n",
    "    else:\n",
    "        b += 1\n",
    "    print(clf.predict_proba(x))\n",
    "    print(clf.score(x, y))\n",
    "    print('===')\n",
    "print(g, b, g/(g+b))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00878202 0.00286738 0.00104342 0.00174449 0.97678358 0.00256539\n",
      "  0.00621372]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.50586901 0.31187084 0.05267682 0.03981707 0.03696206 0.04212153\n",
      "  0.01068267]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.06447345 0.01834486 0.00420684 0.00858639 0.84605413 0.05245723\n",
      "  0.0058771 ]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.04085003 0.02778818 0.01278251 0.00696172 0.15826548 0.73428459\n",
      "  0.01906751]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[3.19742185e-04 7.73776117e-04 1.06130336e-03 6.12729959e-03\n",
      "  9.91279888e-01 3.15434170e-04 1.22556744e-04]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.0106737  0.0077381  0.04230071 0.88822384 0.0418365  0.00539609\n",
      "  0.00383106]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00151725 0.00337314 0.00341552 0.00558851 0.97842565 0.00571792\n",
      "  0.001962  ]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.00776338 0.01129805 0.13707682 0.7914273  0.04253105 0.00572881\n",
      "  0.00417458]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.01467939 0.00986371 0.00563915 0.00381808 0.08036086 0.01648102\n",
      "  0.86915779]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01898052 0.00823891 0.00777486 0.00667336 0.93933543 0.0135315\n",
      "  0.00546542]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.02696916 0.01957414 0.00942175 0.0051148  0.11962183 0.80570642\n",
      "  0.0135919 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01560289 0.00620738 0.00214848 0.00437227 0.82434798 0.13970289\n",
      "  0.00761812]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[7.59551815e-04 3.39931947e-03 1.47040100e-02 1.90270528e-02\n",
      "  9.52415667e-01 8.02119395e-03 1.67320522e-03]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.00399514 0.00547715 0.77909669 0.08820609 0.1043695  0.01391178\n",
      "  0.00494366]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.02696916 0.01957414 0.00942175 0.0051148  0.11962183 0.80570642\n",
      "  0.0135919 ]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.63693215 0.16909211 0.0277264  0.01683836 0.05657136 0.06056141\n",
      "  0.03227819]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.03789162 0.02625188 0.89260722 0.01410692 0.01846525 0.00579487\n",
      "  0.00488225]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.0630458  0.04134052 0.01554759 0.01478304 0.57339246 0.16956196\n",
      "  0.12232862]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00269174 0.00766187 0.00344371 0.01449592 0.9632722  0.00171634\n",
      "  0.00671822]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.03758774 0.01004225 0.02888124 0.00166117 0.89282362 0.00711023\n",
      "  0.02189377]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.01915164 0.01385731 0.83434028 0.09714513 0.02802744 0.00382579\n",
      "  0.00365241]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[9.86917286e-04 2.20820521e-04 1.62352191e-03 9.53180580e-04\n",
      "  9.95547375e-01 3.99082372e-04 2.69102405e-04]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[9.86917286e-04 2.20820521e-04 1.62352191e-03 9.53180580e-04\n",
      "  9.95547375e-01 3.99082372e-04 2.69102405e-04]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.02264855 0.01664406 0.8630785  0.07243791 0.01684816 0.00433913\n",
      "  0.00400369]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.00823378 0.00591311 0.00331877 0.00213577 0.10204327 0.00809698\n",
      "  0.87025832]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00394743 0.01124157 0.0130716  0.03662435 0.92947313 0.00337502\n",
      "  0.00226689]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.00823378 0.00591311 0.00331877 0.00213577 0.10204327 0.00809698\n",
      "  0.87025832]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[3.43732333e-03 1.61735947e-03 7.68021214e-04 1.43581962e-03\n",
      "  9.76834032e-01 7.12565581e-03 8.78178880e-03]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.69204099 0.10012469 0.01932115 0.01184363 0.12766135 0.02732516\n",
      "  0.02168303]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.00858926 0.00640576 0.06048422 0.89054311 0.0261084  0.00456985\n",
      "  0.0032994 ]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.35954383 0.48009642 0.02564782 0.01755757 0.06364417 0.04334751\n",
      "  0.01016269]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.00858926 0.00640576 0.06048422 0.89054311 0.0261084  0.00456985\n",
      "  0.0032994 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00678614 0.01013194 0.02024392 0.02118394 0.9289035  0.00439474\n",
      "  0.00835584]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.01241186 0.01625055 0.01267514 0.84783959 0.09496512 0.01250773\n",
      "  0.00335001]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.32205514 0.4160347  0.02045997 0.04537927 0.12450277 0.05660972\n",
      "  0.01495842]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.01467939 0.00986371 0.00563915 0.00381808 0.08036086 0.01648102\n",
      "  0.86915779]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[9.47406732e-04 2.04637275e-03 4.22797290e-03 4.39673009e-03\n",
      "  9.80484311e-01 1.99451839e-03 5.90268813e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00705537 0.01169389 0.00992496 0.01720278 0.94457276 0.00644546\n",
      "  0.00310478]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01730701 0.00693867 0.003279   0.00143519 0.93633929 0.02831707\n",
      "  0.00638377]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.01467939 0.00986371 0.00563915 0.00381808 0.08036086 0.01648102\n",
      "  0.86915779]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.03841259 0.02958376 0.77959929 0.07903815 0.03962977 0.02098464\n",
      "  0.0127518 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00235359 0.00273424 0.01902502 0.01240163 0.95813286 0.00152153\n",
      "  0.00383112]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00260478 0.0077366  0.00750162 0.01752517 0.95570762 0.00175253\n",
      "  0.00717167]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00403415 0.00151289 0.00244754 0.00109301 0.97439395 0.01514427\n",
      "  0.00137419]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.02589527 0.01768724 0.00529877 0.01123982 0.29097365 0.63744925\n",
      "  0.011456  ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.04738163 0.02991667 0.01131123 0.00943862 0.71192336 0.0302374\n",
      "  0.1597911 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00464125 0.00731573 0.01219629 0.01339825 0.95071675 0.00236008\n",
      "  0.00937165]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[3.19742185e-04 7.73776117e-04 1.06130336e-03 6.12729959e-03\n",
      "  9.91279888e-01 3.15434170e-04 1.22556744e-04]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.03884433 0.02366759 0.00856878 0.00652685 0.06668886 0.83901119\n",
      "  0.01669239]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[6.81142828e-04 7.60749721e-04 3.10877237e-03 5.78689329e-03\n",
      "  9.86141675e-01 6.57729258e-04 2.86303798e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.0453577  0.01543102 0.00616936 0.01000554 0.82123108 0.08289341\n",
      "  0.01891188]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.04738163 0.02991667 0.01131123 0.00943862 0.71192336 0.0302374\n",
      "  0.1597911 ]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.33526225 0.42129443 0.01400792 0.03746925 0.15034247 0.03330525\n",
      "  0.00831843]]\n",
      "0.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.25943161 0.57638271 0.02838741 0.03420375 0.04189771 0.03610004\n",
      "  0.02359678]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.01467939 0.00986371 0.00563915 0.00381808 0.08036086 0.01648102\n",
      "  0.86915779]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00269174 0.00766187 0.00344371 0.01449592 0.9632722  0.00171634\n",
      "  0.00671822]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.00823378 0.00591311 0.00331877 0.00213577 0.10204327 0.00809698\n",
      "  0.87025832]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.03884433 0.02366759 0.00856878 0.00652685 0.06668886 0.83901119\n",
      "  0.01669239]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) R']]\n",
      "[['BPPV (HC) R']] [1.]\n",
      "[[0.40324605 0.46405496 0.0267177  0.01797752 0.04936014 0.02407304\n",
      "  0.01457059]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.04704743 0.05029671 0.02011218 0.82155823 0.04484698 0.01204989\n",
      "  0.00408858]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis R']]\n",
      "[['Vestibular Neuritis R']] [6.]\n",
      "[[0.00823378 0.00591311 0.00331877 0.00213577 0.10204327 0.00809698\n",
      "  0.87025832]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[6.68027156e-03 1.99183873e-03 6.79922633e-04 1.25934695e-03\n",
      "  9.82974343e-01 1.91620807e-03 4.49806925e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00394743 0.01124157 0.0130716  0.03662435 0.92947313 0.00337502\n",
      "  0.00226689]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[1.19056270e-03 2.65891094e-04 1.98634746e-03 1.54999219e-03\n",
      "  9.93486516e-01 5.00617147e-04 1.02007348e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[1.04470824e-03 3.33264046e-03 4.80644073e-03 1.83936466e-02\n",
      "  9.70347325e-01 8.59688823e-04 1.21554999e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[3.66559706e-03 1.85217599e-03 1.38442407e-03 5.96726915e-04\n",
      "  9.73215885e-01 8.56173649e-03 1.07234544e-02]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[3.66988897e-03 1.52129868e-03 6.49710613e-04 8.75735216e-04\n",
      "  9.81644815e-01 7.66787194e-03 3.97067957e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[2.09634491e-03 7.79327402e-04 6.56726224e-04 1.42610750e-03\n",
      "  9.87033608e-01 7.27012011e-03 7.37765479e-04]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01403673 0.01133935 0.00466065 0.00908201 0.84838669 0.08000093\n",
      "  0.03249364]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.00769671 0.00580411 0.07719128 0.87653467 0.02529155 0.00431881\n",
      "  0.00316286]]\n",
      "1.0\n",
      "===\n",
      "[['Vestibular Neuritis L']]\n",
      "[['Vestibular Neuritis L']] [5.]\n",
      "[[0.02696916 0.01957414 0.00942175 0.0051148  0.11962183 0.80570642\n",
      "  0.0135919 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[6.81142828e-04 7.60749721e-04 3.10877237e-03 5.78689329e-03\n",
      "  9.86141675e-01 6.57729258e-04 2.86303798e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[2.94114151e-04 5.98606092e-04 1.40253669e-03 5.29325985e-03\n",
      "  9.90405542e-01 3.76160893e-04 1.62977991e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00668146 0.0122971  0.01459732 0.02768051 0.91739169 0.01484884\n",
      "  0.00650307]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.04860971 0.01321851 0.00570914 0.00874954 0.87484227 0.04387009\n",
      "  0.00500074]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.00878202 0.00286738 0.00104342 0.00174449 0.97678358 0.00256539\n",
      "  0.00621372]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[5.97373530e-04 6.14947209e-04 9.87005233e-04 3.38031469e-03\n",
      "  9.91846664e-01 4.94920342e-04 2.07877549e-03]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[2.94114151e-04 5.98606092e-04 1.40253669e-03 5.29325985e-03\n",
      "  9.90405542e-01 3.76160893e-04 1.62977991e-03]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) L']]\n",
      "[['BPPV (PC) L']] [2.]\n",
      "[[0.03841259 0.02958376 0.77959929 0.07903815 0.03962977 0.02098464\n",
      "  0.0127518 ]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01702649 0.0047343  0.00364684 0.00420919 0.95144166 0.01688235\n",
      "  0.00205916]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.63693215 0.16909211 0.0277264  0.01683836 0.05657136 0.06056141\n",
      "  0.03227819]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[0.01671844 0.00746667 0.00386532 0.00170363 0.87085366 0.09227314\n",
      "  0.00711915]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (PC) R']]\n",
      "[['BPPV (PC) R']] [3.]\n",
      "[[0.02286939 0.02504598 0.02183054 0.83568631 0.08893306 0.00177334\n",
      "  0.00386139]]\n",
      "1.0\n",
      "===\n",
      "[['BPPV (HC) L']]\n",
      "[['BPPV (HC) L']] [0.]\n",
      "[[0.69204099 0.10012469 0.01932115 0.01184363 0.12766135 0.02732516\n",
      "  0.02168303]]\n",
      "1.0\n",
      "===\n",
      "[['Stroke']]\n",
      "[['Stroke']] [4.]\n",
      "[[5.97373530e-04 6.14947209e-04 9.87005233e-04 3.38031469e-03\n",
      "  9.91846664e-01 4.94920342e-04 2.07877549e-03]]\n",
      "1.0\n",
      "===\n",
      "84 1 0.9882352941176471\n"
     ]
    }
   ],
   "source": [
    "g = 0\n",
    "b = 0\n",
    "for x, y in zip(encoded_train, y_train.tolist()):\n",
    "    x = x.reshape(1, -1)\n",
    "    y = np.array([[y]])\n",
    "    print(encoder.inverse_transform(y))\n",
    "    pred_indx = clf.predict(x)\n",
    "    y_est = pred_indx.reshape((1, 1))\n",
    "    print(encoder.inverse_transform(y_est), pred_indx)\n",
    "    if y == y_est:\n",
    "        g += 1\n",
    "    else:\n",
    "        b += 1\n",
    "    print(clf.predict_proba(x))\n",
    "    print(clf.score(x, y))\n",
    "    print('===')\n",
    "print(g, b, g/(g+b))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.score(encoded_test, y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0.9882352941176471"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "clf.score(encoded_train, y_train)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# All in one"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 127,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data has 95 rows and 54 columns.\n",
      "Creating mask to determine which rows can be dropped\n",
      "Currently 0 rows are being dropped\n",
      "=====================\n",
      "Checking for entries that are droppable by being rare...\n",
      "0 rows are being dropped\n",
      "=======================\n",
      "Checking for unknown values that require warnings...\n",
      "=======================\n",
      "Checking for columns that lack signal...\n",
      "=======================\n",
      "Data has 95 rows and 54 columns.\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "581c68126b584e869dbac7ad523ec8bd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/30 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "7\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['BPPV (HC) R']] [['BPPV (HC) L']]\n",
      "18 1 0.9473684210526315\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e8488516a56a4f3b8168f1f4045c1f0e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/30 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "58\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['BPPV (HC) L']] [['BPPV (HC) R']]\n",
      "18 1 0.9473684210526315\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5e2a943f0b0a490f950088cfe85a7264",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/30 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "20\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['BPPV (HC) L']] [['BPPV (HC) R']]\n",
      "18 1 0.9473684210526315\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f6d67805f62f43f7b1b1ab7c49f1eb55",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/30 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "59\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYAAAAEGCAYAAABsLkJ6AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/YYfK9AAAACXBIWXMAAAsTAAALEwEAmpwYAABCA0lEQVR4nO3dd3hc5ZU/8O+ZLk1RlyxblnuRsbENbqFDCAFCIAmBwAKbZAubTfnBlrTdbBI2u5tKskkIbNiUJaGFEEggoYQAxhDA3cbGki3b2OoatRmNyvT398edOxrJU+6UO3fK+TyPH1TnvoNGOvd93/OeQ0IIMMYYKz86rQfAGGNMGxwAGGOsTHEAYIyxMsUBgDHGyhQHAMYYK1MGrQegRH19vVi8eLHWw2CMsaKyd+/eYSFEQ6LPF0UAWLx4Mfbs2aP1MBhjrKgQ0elkn+clIMYYK1McABhjrExxAGCMsTLFAYAxxsoUBwDGGCtTHAAYY6xMcQBgjLEyVdIB4MX2Qdy7/bjWw2CMsYJU0gHg1c5h3PvyCa2HwRhjBamkA0CD3YwJXxDeQEjroTDGWMEp6QBQZzUBAIYnfBqPhDHGCo9qAYCIfkZETiI6HPOxWiJ6gYg6I/+tUev6AFBvMwMARib8al6GMcaKkpozgP8DcOWcj30BwItCiBUAXoy8r5o6G88AGGMsEdUCgBBiB4DROR++DsADkbcfAPABta4P8AyAMcaSyfceQJMQoh8AIv9tTPSFRHQ7Ee0hoj1DQ0MZXUwOAMOTPANgjLG5CnYTWAhxvxBikxBiU0NDwn4GSVWY9LCa9Bj28AyAMcbmyncAGCSiZgCI/Nep9gXrbGaM8AyAMcbOkO8A8BSAj0be/iiA36l9wTqbiTeBGWMsDjXTQB8B8AaAVUTUQ0R/DeAbAN5DRJ0A3hN5X1X1NjNvAjPGWByq9QQWQtyc4FPvVuua8dTbTNjfNZbPSzLGWFEo2E3gXKm3mTE66UcoLLQeCmOMFZSSDwB1VhPCAhib4mUgxhiLVfIBoN7Oh8EYYyyekg8AdVY5AHAmEGOMxSr5ANBgl+oBDXEAYIyxWUo+AMzMAHgJiDHGYpV8AKiqMMKgIz4Mxhhjc5R8ANDpCLVWE88AGGNsjpIPAIBUD4hnAIwxNltZBIB6mwnDkzwDYIyxWGUSAMwY9vAMgDHGYpVJADBhZNIHIbgcBGOMycoiANTZzPAGwpjyh7QeCmOMFYyyCADR1pC8EcwYY1FlEQDqbNJp4GFOBWWMsaiyCAANPANgjLEzlEUAkGcAfBiMMcZmlEcAsPIMgDHG5iqLAGAy6OCwGLgkNGOMxSiLAABIjWF4E5gxxmaUTwCwcj0gVvoe2nka33i2Q+thsCJRNgGgzmbiAMBK3jOH+vG7A71aD4MVibIJAPU2M0a4IBwrcc5xH0Ym/Vz2hClSNgGgzmaCayqAQCis9VAYU43T44M/GMaEL6j1UFgRKJsAIJeDGOVZACtR3kAI7ukAAH6dM2XKKADI5SB4H4CVpqGYkue83MmUKKMAIB8G418MVpqcMQFglF/nTIGyCQB1kQDAh8FYqRryeKNvj0zy65ylVjYBgJeAWKnjJSCWLoPWA8gXm9kAk0HHBeFYyXJ6fNARYNTreAmIKaLJDICI/oGI3iaiw0T0CBFZ8nBNNNjMGOIZACtRznEf6mxm1NvMnAXEFMl7ACCiBQD+H4BNQoi1APQAbsrHtetsJp4BsJLl9HjRaDdLr3MOAEwBrZaADAAqiCgAoBJAXz4uWmc18QyAlSynx4dGuxkCvNfFlMn7DEAI0QvgOwC6APQDcAsh/jj364jodiLaQ0R7hoaGcnLtepsZwx6+M2KlSQoAFtRaTbwHwBTRYgmoBsB1AJYAmA/ASkS3zv06IcT9QohNQohNDQ0NObl2nc2MkUkf10lhJScUFhiZ8KHRYY7WveLXOUtFi03gywG8I4QYEkIEADwB4Lx8XLjeZkIgJDA+zXVSWGkZmfAhLIBGuxm1VhN8wTCm/CGth8UKnBYBoAvANiKqJCIC8G4A7fm4cPQ0MB+SYSVGPgXcEFkCArgHNktNiz2AnQAeB7APwKHIGO7Px7WjAcDDAYCVFmfkFHCjw4w6OQDwjQ5LQZMsICHEVwB8Jd/XrbPJvxh8Z8RKi3Nc+mPfaDdDRwSAK4Ky1MrmJDAQWxCO74xYaRmKLgGZIe/98o0OS6VsagEBQE2lEURcEZSVHqfHh6oKI8wGfXQPgGcALJWyCgAGvQ61lSauCMpKjnwKGAAqTXqYDTp+nbOUyioAANwcnpUmp0c6AwBIda+4BzZTovwCgNXM6XGs5DjHpVPAslqriZeAWEplFwDq7WaeAbCSIoTAUKQOkIwDAFOi7AJAnZUrgrLS4p4OwB8KoyEmAPDrnClRdgGgwW6GxxeEN8DH5FlpkE8BNzpmLwHxQTCWStkFgJlTknx3xEpD7CEwWZ3NDG8gjCk/171iiZVdAOByEKzURMtAzFkCArgeEEsuYQAgotUxb5vnfG6bmoNS00w5CA4ArDQkWgIC+DAYSy7ZDODhmLffmPO5e1UYS17MlIPgXwxWGpzjPlSa9LCZZyq71PKNDlMgWQCgBG/He79ocD0gVmqGJmangAJAvVV6n5eAWDLJAoBI8Ha894tGhUkPq0nPvxisZDjHvbNSQIGZGQAvAbFkklUDbSGiH0C625ffRuT9BaqPTEV1Nj4MxkrHkMeHtmbHrI9ZTXqYDDoOACypZAHgszFv75nzubnvF5V6Gx+SYaXD6fHhopWzZwBEhDqrife6WFIJA4AQ4oG5H4s0dHeJIu82XWczo3t0SuthMJa1KX8QE75gtBBcLKkcRPYz3f1dY3BNB3DpqsasH4sVlmRpoF+WU0GJyExELwE4AWCQiC7P1wDVUG/jOyNWGmYOgVnO+FydzZyTJaDv/akTdz31dtaPwwpPsk3gjwA4Gnn7o5DW/hsAXAzgv1Qel6rqbWaMTvoQChf1RIaxmTMA9jNnAHVWU05OvPe5pjE47kORT/xZHMkCgD9mqee9AB4VQoSEEO0o8laSdVYTwgJwTfEsgBW32Gbwc+WiIqgQAv2uaUwHQvD4uKxEqUkWAHxEtJaIGgBcCuCPMZ+rVHdY6qq382EwVhqSLQHVWk2Y8ocw7c+88OH4dBCTke+Xr8VKR7IAcCeAxwF0APieEOIdACCiqwHsV39o6qmLHpLhFzQrbk6PD0Y9oabSeMbn6nNwGrjPPT1zrXFvxo/DClOyLKA3AayO8/FnADyj5qDU1mCXfjGGOUeaFTmnx4sGmxlEZx7Or43c6IxO+tFSk9mkvc8VEwC4gGLJSRgAiOgfk32jEOK7uR9OfsgzAK4IyordkMeHBseZyz/ATEG4bDaC+9wzd/2DPAMoOck2c78D4ACAZwH4UMT1f+aqqjDCoCMulMWK3pDHl/DuPhcloftc0zDoCEa9DoO8B1BykgWAcwDcBOB9APYCeATAi8V+CAwAdDpCrdWEYQ8vAbHi5vT4cM6imrifq4vWA8r8D3e/axrzqiww6CiaccRKR8JNYCHEASHEF4QQGwD8FMB1AI4Q0bX5Gpya6m1mngGwouYPhjE66Y97BgAAbGYDTHpd1ktA86sr0OiwcBZQCUrZESySBroRwDoAPQCcag8qH+psJgxxGigrYnJBw3gpoIBUD6jWasJolktA86ssaLSbeQZQgpJtAn8c0mlgC6R00BuFECXxxx+QZgDvDE9qPQzGMpbsFLAsm8NgobDA4Lg0A/AFw3ix3QkhRNyMI1acku0B/BTAIQBdkE4CXxH7gxdCZLwURETVAH4CYC2k3gJ/JYSY23VMVVI9IB+/oFnRkvPy450CltXZTBmnOw9P+BAICTRXV2DKF8R0IIQJXxB2y5lnDlhxShYALlXxut8H8JwQ4sNEZIIGJ4vrbGZ4A2FM+UOwmou6sgUrUzMzgPhLQICUCXRqJLOZrnwGYH6VBR6vVAZicNzHAaCEJDsI9ooaFyQiB4CLAHwsch0/gLwvxsutIUcm/BwAWFFyenwgmjnxG0+t1ZzxHkCfS5phzK+uwFikbpbT48XyRltGj8cKT8pNYBUsBTAE4OdEtJ+IfkJE1nwPQk6RG+JyEKxIDXm8qLOaYNAn/jWus5kw6Q/BG0i/HlC/W54BVERnGZwJVFq0CAAGSGcM7hNCbAQwCeALc7+IiG4noj1EtGdoaCjng2iwqVsPSAiBB14/xfWGmGqc4z40JFn+AbI7DdzrmobVpIejwoCmyD4DZwKVFi0CQA+AHiHEzsj7j0MKCLMIIe4XQmwSQmxqaGjI+SDkGYBaFUG7RqfwlafexqO7u1V5fMaGJnxJM4CAmdPAmSwD9bu8aK6uABHBZjagwqjn08AlJlka6NOQMnTiyjQLSAgxQETdRLRKCHEUwLsBHMnksbKhdkXQ3sgGWnv/uCqPz5hz3IdVTfakX1OXRUXQfvc0mqukGQYRoclh5oJwJSZVLSAA+BCAeQAejLx/M4BTWV73MwAeimQAnQTw8SwfL20mgw4OiyF6mCbX+iMbaEc4ADAVhMMCwxM+NKSYAcRWBE1Xr8uLtmZH9P1Gu4ULwpWYlFlARPQ1IcRFMZ96moh2ZHNRIcQBAJuyeYxcqLebVSsJLW+gnRqexLQ/hAqTXpXrsPI0OuVHMCxSLgHJewDpBgBfMIThCR+aqyqiH2t0mPF2H9/QlBIlewANRLRUfoeIlkDqDVz06q1m1UpCy2V0wwI4OuhR5RqsfEU7gSUoBS1zWAww6intva4Bt5wCOvP4PAMoPUoCwD8A2E5E24loO4CXIXULK3p1ttw0zY6n3zUd7dLE+wAs16K9gFPMAKL1gNLcA4g9AyBrcpgx5ZdOA7PSkPIElBDiOSJagZnuYB1CiJLYCaq3mfHmyRFVHrvf7cW5i2qw8+QojvC0meWYklPAslqrOe0loOgp4OrZS0CA1BjG1sCHwUqBkmqglQA+C+DTQoiDAFqJ6BrVR5YHdTYTxqYCCITCOX/sPtc05ldXYHWznWcALOeG5ACQpA6QrM6a/kxX3sOSs4AAoCkSbHgZqHQoWQL6OaRSDe+KvN8D4D9UG1EeyeUgxnK8DDThC2LcG0RzVQXamh3oGPAgHC76PjqsgDjHvbBbDLAYUycX1FpNaXcF63N7UWs1zXp8OdgMcSpoyVASAJYJIb4FIAAAQohplEh7yHqVykH0R6fPFqxpdmDCF0TP2HSK72JMOacn9SEwWZ0t/ZLQ0gx29vKSvOHMM4DSoSQA+ImoApFDYUS0DFKP4KIXWxAul+QMIHkGAPB5AJZbUgBIvf4PSEtAE74gfEHl9YD6Xd5ZKaAAYDcbYDHquB5QCVESAL4C4DkAC4noIQAvAvicqqPKk7pIAMj1YTB5BtBcZcGqeXboiAMAy60hj0/R+j+Q2WGwPtc0FlTPDgDSaWALBnkJqGQoyQJ6gYj2AdgGaennDiHEsOojywN5CUiNGQARMK/KAqNehyX1Vt4IZjkjhIDT41W8BBQtCDfhP+OuPp5xbwAeX3DWBrCsyW6JNqJhxU9JFtD5ALxCiD8AqAbwL0S0SO2B5YPNbIDJoFNlBtBoN8MYKdPb1uzgAMByxuMLwhsIpywDIZupB6TsRqc/zhkAWQPXAyopSpaA7gMwRUTrIaWDngbwC1VHlSdEhAabOecVQfvds9dP25od6Bmbxrg3kNPrsPIUPQWcxh4AAMWHwfrcM0kMc/EMoLQoCQBBIYQAcB2AHwghvg8geQnCIiKdBs7tHU2fe3YGxZr50kZwRz+XhGDZU3oKWDZT+Ta9GUC85aJGhxmTfBq4ZCgJAB4i+iKAWwH8gYj0AEqmKWi9zZzTJSAhxBkZFGvkTKA+d86uw8pXOofAAMBRYYBBR4o3gftc09DrKG6AiTaG4VlASVASAD4CKe3zr4UQAwAWAPi2qqPKo7oMDskk454OYDoQmrWB1mg3o9ZqQjvPAFgOyEtAqbqByYgINWm8zvvc02iym+O2mmyMngbmfYBSoCQLaADAd2Pe70KJ7AEAUiroyIQfQggQZX++rTdODRUiQluzHe0DvBHMsuf0eGGO9LNQKp1yEHIZk3i4NWRpSTgDIKLXIv/1ENF4zD8PEZXMX7J6mwn+UBjj3tysac6sn86+O2ub58DRAQ+CKtQdYuXFGTkDkM4Ni3QaWNlde79bagUZTwM3hy8pCQOAEOKCyH/tQghHzD+7EMKR6PuKTX2OD4P1u8+cAQDSRrAvGMapkcmcXIeVL+e48lPAMqUVQcNhaQ8rXgYQIPUXsBh1PAMoEclmALXJ/uVzkGrKdTmIPrcXRr2UXhpLLgnBHZVYttI5BCZTugQ0MumHPxTG/AQHxogo0hiGZwClINki4l5I9X/izTMFgKVxPl505EMyOZsBuKbR5LBAp5v9v21Zgw1GPaG934PrNuTkUqxMOT0+XLC8Pq3vqbOa4PFK9YDMhsQVROOVgZ5Lag7PM4BSkKwn8JJ8DkQrMzOA3ASAPrc37t2TyaDD8kbuDcCy4w2E4PEGU7aCnKs2cqMzNhnAvKrEASBeI5i5Gu0Wfh3nwciEDz986Thu2dqKFU3qHL1SkgYKIqohoi1EdJH8T5XRaKCm0ggi5Ow0cL97Gs0J1k/buDkMy5J8BmDuEmMq8mngVIce47WCnKvRYeaS0Hmwv8uF/3v9FFzT6lUQUFIL6G8A7ADwPIC7Iv/9qmojyjODXofaSlNOloDCYYEB95lldGVrmh1wenw5rz3Eyoe89NKg8BCYTGlF0D7XNCxGXbSfdTxNDgufBs6Dgz0u6HWEs+arl3OjZAZwB4DNAE4LIS4FsBHAkGoj0kCdLTeHwYYnfQiERMIMCvlEMM8CWKZm6gClGwDkekDJX+f9kSXMZCmm8rX5NLC6DnS7sLLJjkqT8vMe6VISALxCCC8AEJFZCNEBYJVqI9JArspB9CWpoQLMZAJxAGCZSqcZfKz6aLJD8gDQ60q8hClriuw/cFVQ9YTDAge7XdiwsFrV6ygJAD1EVA3gtwBeIKLfAehTc1D5Vmczp900O57YRjDx1FhNmOewcEkIljGnxwu9jqJr+ko5LEbodZTyMFi/ezphCqhMngHwPoB6To1MYtwbxIaFVapeR0kpiA9G3vwqEb0MoApSh7CSUWc1YTgHdzNyK8hkG2i8Ecyy4Rz3od5mOiPNOBWdjlBTmbw3cCAUhtPjS3gKWCZnIHFzePUc6HYBANYXwAxAzgI6G4AHQA+AtaqOKs8a7OZIkw3lPVPj6VewgdbW7MBx50Ra/VljeQMhSNW5WTlKpxfwXKkKHw64vRACWJBiCchhMcBs0PEMQEUHu12oNOmxolHdyvtKsoC+BuAtAD8EcHfk33dUHVWezaTIZbcMpGQDbc18B4Jhgc7BibQfv3t0Cud87QU8//ZgNsNkRUwKAOltAMuk3heJX+P97uR7WDK5NzDvAajnQI8b6xZUQZ/mTC9dSmYANwJYJoS4WAhxaeTfZaqOKs9ydRisL8kZAFk2G8EP7+rClD+EPadGMxofK35DHq/iPgBz1VqTLwEpOQQma7TzWQC1+IIhtPeNY0NrterXUhIADkPqBVyyclUOYm4jmHgW11lhMerS3gj2BUN4bHc3AOCYM/3ZAyt+wVAYI5N+xX0A5pKWgBK/xpO1gpyLZwDqae/3wB8KY0NLterXUpJg+nUA+4noMKTGMAAAIcS12Vw40llsD4BeIcQ12TxWtqIVQT2ZLwEFQ2E4PV7MT1JDBQD0OsKqeek3iX/u8ABGJv1YUF2BzkHOIipHI5N+CJH+GQBZrdWMcW8QgVAYxjjNXvpc06iqMCrKO2+wm/HKMQ4AajiYpw1gQFkAeADANwEcApDLYvZ3AGgHoHlp6XlVFpj0OpwczrxU86DHh7BAygwKAFjTbMczhwbSakLz4JunsbiuEjdsWohvP38U494AHJaS6czJFMj0EJhsph6QP24tIakMdOrXLyDNACZ8QUz6grCa1TuoVI4OdLvQaDcnLciXK0qWgIaFED8QQrwshHhF/pfNRYmoBcD7APwkm8fJFaNeh+WNtqzSM1OdAYi1ptkB93QguumWSsfAOHafGsMtWxdhVaQoVCabyKy4RctAZBgA6q3JD4NJhQyV/dGJngbmZaCcO9jtwvqF1TnpUJiKkgCwl4i+TkTvIqJz5H9ZXve/AXwOSWYURHQ7Ee0hoj1DQ+pXnlidZX5+vFaQiaS7Efzgm6dhNujw4XNbsDIaAHgZqNxETwGnWQlUlqocRLJWkHNFTwPzRnBOuacCODk8qfoJYJmSALARwDYA/4UcpIES0TUAnEKIvcm+TghxvxBikxBiU0NDQ6aXU0wu1JZpJtBMCl3qX87VaQSACV8QT+7rxTVnz0eN1YSWmgpUGPU4xjOAshNtBp9mJVCZnOwQryLopC8I93QgZRabTO4NPFhAM4D/3XES//VMu9bDyMrBHhcA5C0AJF28i2zUPiWE+F4Or3k+gGuJ6GoAFgAOInpQCHFrDq+RttXzpD/KHQMenL88/V+wftc07BYD7ArW5W1mA1prKxVlAj25vxeT/hBu3dYKQDrRubzRhk4nzwDKjdPjRU2lESaDovObZ0hWEVRuBLNA4Qyg0V54M4An9veia2QSn79yter582qRN4DXtahbAkKW9JUkhAgByCrbJ85jflEI0SKEWAzgJgAvaf3HH5BKNACZF2pL1Agm2fWOpLiWEAIPvnEaaxc4Zt0RrGiy4RgvAZWdbE4BA0B1hRE6it/+NFUhw7kcFdJp4ELZAwiFBU4MTWDSHyrqm6ODPS4sa7DmLcFDya3E60R0DxFdmMM9gIJTZzOj0W7OuFBbskYw8bQ1O3BqZBJT/sQ11fecHsPRQQ9u3bpo1obQyiY7Bsd9cKvYKIIVHqfHl/EhMECaPdYm6A3cl0YSAxDpDVxAjWG6RqfgD0pbivu7XNoOJkNCCBzodmHDwpq8XVNJADgPwFkA/h05LgUhhNiu9RmAWKub08/Plyk5BBZrTbMDQkhLTon88o3TsFsMuHbD/FkfX9lkA8AbweVmaNybcQaQTDoNfOZde5/bCyIpJVqpJrslui+htdgZ8f6uMQ1Hkrle1zSGJ/yqVwCNpaQa6KX5GEghaGu24+cnRhIelEnEGwhhZNKvOIVOutbMRvA5rWdG/OEJH5493I9bti4642COXCDq2OAENi2uVXxNVryEEBiayG4JCEhcDqLfNY1Guzmt132jw5z0Biaf5JuhLUtqo5U0i83BbjcAFNYMgIiqiOi7ckomEd1NRPkLUXm0ptkBfyiMk0PpHQgbkDOAFG6gAUBLTQXsFkPCGcevdncjEBK4dduiMz63oLoClSY97wOUkbGpAAIhkfEhMFmdNX7viz638hRQWaPdgqECmQF0OiewoLoC5y+rR6dzAuPe4lsePdA9BpNBh1Xz1K0AGktJuP8ZpDLQN0b+jQP4uZqD0oqcCZTuMlC0hkoaMwAiQts8B470nXmtUFjg4Z1deNfSOixvtJ3xeZ2OsIIzgcrKUPQMQJYBIEH7035XekkM8lg8vmDSfax8OTY4gRVNNmxsrYYQwFuRu+licrDbjbPmOzLO8sqEkistE0J8RQhxMvLvLgBL1R6YFpY2WGHS69IPAK70ZwCAtOTUMeBBODy7vv/2o070uqZx27vOvPuXrWiy81mABLyBENxTxXcHmIx8CjgXS0Du6QACoZkzmEII9LqmFRWBi9UUTQXVdhYgZwCtbLJH6+cc6C6ufYBgKIxDve685f/LlASAaSK6QH6HiM4HMK3ekLRj1OuwosmG9jTXNdMpAxFrzXwHpvwhdI1Ozfr4g2+eRqPdjPesaUr4vSubbBjy+OCayr6VZam56+m3cc09ryIYymXpKm1FD4FlvQQUqQcU87oZmwrAFwynlcQAzMxGtM4EkjOAVjTaUFVhxLIGa9FlAh0bnMB0IFSQAeATAH5ERKeI6DSAeyIfK0mrM6jU2ef2os5qgsWoT+v74pWE6B6dwvZjQ7hpS2vSDbkVTTMbwWy2vafH0D06jVeOqV9CJF9mmsFnmwV05mGwmT4Aac4ACqQ5vLwXJv9ObGytwYFuV1F1zpNPAK/PQwnoWCkDgBDioBBiPYCzAawTQmwUQhxUf2jaaGu2Y8jjS6s3QLpnAGQrm+zQEWYdCHtoZxd0RLh5y8KU3wuAN4Ln8AVDOBHZxH9kV7fGo8kdp8cLq0mfdeXNaDmIiXgBIN1N4MKYARyP9MdYEdkv29hajZFJP7pHi2eh4kCXC9WVRiyqq8zrdVO+mojIDOB6AIsBGOQDSUKIf1d1ZBpZE7kr7+j34IIVyu62+l1etGbwg7MY9VjaMFOF1BcM4bE93bi8rTHldHx+lQU2s4HPAszROTiBUFhgaYMVLx91YsDtTSu3vVBJh8Cyfx7x2p8qbQU5V1WFVJZC6+bwxwY9WFBdEQ2O8jLK/u6xjH4vtXCwx4X1LfmpABpLyRLQ7wBcByAIYDLmX0lKp1CbrM89nVYGUKy2Zkf09PGzhwYwOumPm/o5F5FUE4iXgGaT89L/7X1rEAoL/HpPacwChsZ9Wa//AzEVQWNmuH2uaZgMumhwUErqDaz9aWA5A0i2qsmOCqO+aPYBJn1BHBv05KUBzFxK5pMtQogrVR9Jgai1mtDkMCsOABO+IDzeYNoZQLI1zQ48fbAP7qlAtOnL+cvqFX3vyiYbXmx3ZnTdUtXePw6LUYeLVjbg/OV1eHR3Nz516XLoirQ4mMzp8WLtguyP31RXmkA0Zw/A7UVzlSWj/0eNdm1bQ8oZQBeumPmdMeh1WNdShf1FciDsUK8bYQFs1CAAKK0FtE71kRSQtmaH4kygTDOAZq4lreU/ub8He06P4dZtixT/Iq5ssmNk0p91M/tS0t4/jlVNduh1hJs2t6LXNY1Xjw9rPaysZVsITqbXEWorZ9cD6nNNZ/z61XoGIGcAzT0vs7G1Gkf63PAGQhqNTDm5AujZeaoAGktJALgAUlOYo0T0FhEdIqK31B6YllbPc+C40xMtLpVMX2T9NN0NNJm853D3C8eiTV+UWsmZQLMIIdDePx7NrrrirCbUVBrx6K4ujUeWnQlfEFP+UNaHwGS11tmHwfrTaAQzl9YzADkJQv5dkG1cWINASKSsuFsIDva40FpbiboM+zxkQ0kAuArACgBXAHg/gGsi/y1Zbc12BELS1DKVTDMoZA12M+qsJni8Qbx//XxUVypfh412B+MTwQCku+SxqQBWR47Smw16fPjcFrxwZFDzjcpsnIr0qm7KYQCQl4CCoTAGxtM/BSxrdJjh8Wp3GljOAIo3AwCKozLogS6XJuv/gLI00NPx/uVjcFqJZgINpL576HdNQ0dAU4YbdEQUvWNVsvkbq8lhht1i4FTQCHnfRv7/CQAf2dyKYFjg8b09Wg0raw+8fgpmgw4XrshNZ7w6mynaFczp8SEsMr+BadT4NLCcAWSbkx7b5LBgfpWl4AvDOce96HN7sV6D5R9A2Qyg7Cypt8Jk0CnqDdDn9qLRboEhjSqKc11/7gLcuKkl7RcBEWEll4SIkn9eq2MCwPJGG7YsqcWvdnedUXKjGPS6pvHk/l7cvKUV9TlaIoidAcidwDI5xwLMzEq0WgaamwEUa2NrTcGXhj7YI9Uskmcs+cYBIA6DXoeVTTZFmUCZHgKL9cGNLfjWh9dnlAO8ssmGzkFPUZ16VEt7/zgWVFegqmJ2N6WbtyzEqZEpvHlyRKORZe5/d5wEAPztRbkrv1VnNWNsKoBgKIzeSB0rpa0g55JnAFpsBMfWAIpnw8Jq9IxNF/Ty34HuMeh1hLPm8wygoLQpLAmRSRXFXFrRaMfYVADDcSo8lpuOgfFoVlWsq9Y2o6rCiEd2F9eZgOEJHx7Z1YUPnbMg4z/Q8cingcemAllnsWk5A0iUASST76oLeRnoYLcbq+fZ0y4jkyscABJY3ezA8IQ/6d2DEAJ97sxT6HIhuhFcpPsAY5P+6CZnNrwBqQSEXNI7lsWoxwc3LsDzhwfiNkMpVD977R34Q2F84uJlOX3c6GGwST/6XNOwmw2wZ9iDVj4NrEVz+EQZQLK1C6pg0FHBLgOFwwIHe1x5LwAXiwNAAkqaxLumAvAGwhkfAssFuT1kMW4ET/qCuPHHb+Cm+9/MegnruFMqARG7ARzr5i2t8IfCeGJfcWwGu6cD+OUbp3H1umYsbYh/h5up2mg5CB/63N6MN4CBSG9gu1mTGUCiDCCZxahHW7OjYDOBTg5PwuMNapYBBHAASKhNQXOYTBrB5FqD3YyqCiOOOYtrI1gIgc89/hY6nRMYGPfi1MhU6m9KYiYDKP7d4Kp5dpzTWo1HdnUVxX7JL984BY8viE9ektu7fwDRzWR5BpDtHlajXZvDYIkygGJtbK3GWz0uhAowAUA+AMYzgAJUYzVhnsOStOdpf4aNYHJJygSyFd0S0E9efQd/ONSPGyIH33a9k90GbXu/BxajDovqrAm/5qYtrTgxNIndpwpzSUA25Q/iZ38+hUtXNaiyORidAUz40Z/lDACQUi61mAF0JskAkm1YWI1Jf6ggz8oc7HHBZjZgWY5neOngAJBEW7M96QxATqFLt456rsndwYrhzhYA3jgxgm8814Erz5qHb15/NmqtJux6J7s/yh0D41g1zwF9kjIa15zdDLvZUPAngx/d1Y3RST8+delyVR6/JlIPqM81jdFJf9YzWC1mAHIG0IoEyz+yja1Sg/VCXAY60O3CugVVSV+zauMAkERbswPHnRPwBePXE+l1eWHUE+qt+T/CHWtlow3u6UBBp7vJ+t3T+PTD+7C4rhLfvuFs6HSEzYtrsPvUaMaPGS0BkaKZdqXJgOs2zscfDvUXbMtIfzCM+3ecxJYltdi0uFaVa+h1hOoKI96O9KPOdgbQ6LDA4w1i2p+/ujtdo1PwBcPRJjCJLK6rRHWlEQcKLAB4AyG0949ruv4PcABIanWzA8GwwAln/CyVfvc05mVYRTGXiqUmkC8Ywice3AdvIIQf33ZuNPNk8+JadI1OYcCd2V3k4LhUAiLRBnCsmza3whcM48n9hbkZ/OT+HgyMe1W7+5fVWk041CsdQkq3D8BcM53B8jcL6EyRASQjImxYWI39BdYjuL1/HIGQ0HT9H+AAkNSaFJlA/S5v1r88ubCiSLqD3fX0ERzsduE7N6zH8saZX9wtS6Q73V0ZzgLaB84sAZHI2gVVOLulCo/u7i64JbNQWOC+7SewbkEVLlqhrCR4pupsZrinpVlQtkuYcmewfO4DdKbIAIq1cWENOp0T8HgLZ9Z3oAA2gAEOAEktrrPCbNAlDADZNILJpXqbCTWVxoLc6JI9tqcbD+/swt9dvBRXrWue9bk1zQ5YTfqMN4Lln8+qFEtAsps2t6JjwFNw9eL/cKgfp0am8KlLl6neGSq2+Uu2HdPkGUA+9wGUZADJNrZWQwjgrUjZhUJwsNuFeQ6L5t3qOAAkIZWEsMfNBAqHBQbHvZpmAMmIKLoRXIgO9bjxpd8exnnL6vDZK1ad8XmDXodzFtVgd4YbwR39nrglIBK5dsN8VJr0BbUZLITAvS8fx7IGK65YM0/168mZQPU2M8yG7E6hRmcAeSwIpyQDSCavsxfSgbCDPW6sX6hN+YdYHABSkDOB5i4XDE/4EAiJgpgBANKBsGMFWBNodNKPTzy4F/VWE35488aERfO2LqnF0UEPXFPpn9SVegAou/sHAJvZgGvXz8fTB/sLZlngpQ4nOgY8+OQl+eleJs8AFuQgg6260giTXofBPO0BKM0AklVVGLGswVowJSFcU368Mzyp+QYwwAEgpbZmB0YmzywJ0ZdhI221rGyyw+MNYlCjsrzxhMIC/++R/Rjy+HDfrecmbXixOZLxkm6OvjcQwsnhSUXr/7Fu3tKK6UAIvzvQF/fzQggMuL144cggvvvCMfzV/+3Gf/z+SFrXUEoIgXtePo4F1RW4dsN8Va4xl/yzyMXrl4jQYDdjKE+vPaUZQLE2LKzB/i5XQdwgyRVAtV7/B5T1BC5rcm2Z9gEPGh0zd0v9WTaCybUVjTMbwVqvK8ru/uNRvHZ8GN/40LqUdzvrF1bDpNdh96lRvGdNk+JrpCoBkcjZLVVoa3bg0d1duGVrKwbHfTjU68ahXjcO97rxVo8bw5FWmzoCaq1mvNThxM1bW3N+cOfNk6PY3+XC1647C8YsyoqnQ14CytXrt8lhztsMQGkGUKyNrdX4zb4edI9Oo7WuUq2hKXKgywUiYF0OejxnK+8BgIgWAvgFgHkAwgDuF0J8P9/jUEpuDtPeP46LV8405JhpBVkYf2xjawJdtDI3jUOy8fzbA7h3+wnctHkhbtrSmvLrLUY91i+sws530ssEklv+rVa4ASwjIty8ZSG+/Lu3sfk//xStpqojKbPkopX1WBfJGGprdmDCF8T533gJD73ZhS+/f01a10rl3u3HUW8z44ZNC3P6uMnURQNAbl6/jXYLjivooJcL6WQAyaIdwrrHNA0AQghsP+bEikZbxgX4ckmLGUAQwD8JIfYRkR1Sv+EXhBDqzK+zVFVpxPwqyxmZQH2uaVQY9Yo3HtVWZ5NaS3YWwEawa8qPf3rsINa3VOGr156l+Ps2L67F/TtOYsofRKVJ2Uuzo9+DCqM+aQmIRD64cQFeOTqEqkrjrD/28a5daTLgyrXN+PXebvzze1cqHl8qB7tdeLVzGF+4anVeSwIvrK2EjpDWMkoyTQ4zXj8xnJPHSqUzjQwg2aomOyqMeuzvcuG6DQtUHF1yzx4ewP4uF/7jA2s1G0OsvO8BCCH6hRD7Im97ALQD0O4nosDqZgc65nQHkxvBqJ2ul44VTTYcK4BU0FeODWHCF8RXrj0rrT9qW5bUIhgWaR3bb+8fx8p59oyO09stRvz0Y5vx3Rs34OPnL8G5i2qT/mG/bdsieLxBPH0w/r5BJn708nE4LIa024Fma2FtJV77/GU5O2/Q6LBg3BuEN6D+aeBjgxNp3f0DUqbZupYqTTeCvYEQ/uuZdqyeZ8fNCmbF+aDpJjARLQawEcDOOJ+7nYj2ENGeoaGhvI8tVluzHSeGZpeE6NO4EUw8K5vs6CyAmkA7jg2jutKI9S3VaX3fuYtqoCMoXgYSQqB9YDx6YE9tmxfXYFWTHb9443RO/h8fG/Tgj0cG8bHzl6R1N5sr86srcnYDk69U0JkuYOnvw2xsrcaRvvGEpV3U9pNXT6JnbBpffv8aTev/xNIsABCRDcBvANwphDjjpJUQ4n4hxCYhxKaGBm3XtFfPk0pCxC6v9GvcCCaeFU12TPiC0f0JLQgh8GrnEC5YXp/2i9xuMWLNfAd2KwwAg+M+uKYCcZvAqIGIcOu7FuHtvvGc3Ene89JxVJr0+Ph5i7N+LK3JCRJqbwR3Z5ABJNu4sBr+UDhaAymfBtxe3Lv9BK48ax7OW6buKe90aBIAiMgI6Y//Q0KIJ7QYQzrkDBP5QFggFIbT4yuIQ2CxVjZq3xymY8ADp8eX8Ub05sW12Nc1Bn8wnPJrZ3oA5CcAANK+gdWkxy/fPJ3V4xzudeOpg3346HmLURNzKrdYRVtDqjwDkF/bSs8AxJIrg2pRGO5bz3UgGBL4l6vb8n7tZPIeAEiac/4UQLsQ4rv5vn4mltTPLgkxOO6FENo2gomnENpD7jgmLdddtCKzALBlcS18wXC0UFkycg2g1XlaAgKkQ2QfOqcFv3+rP6v2kt96/iiqKow5b/eolaY8NYeXM4AymQE0OSxorrLkvQTI/q4xPLG/F39z4RLNU1Dn0mIGcD6A2wBcRkQHIv+u1mAciul1hFXzZnoD9Lu1bwQTT43VhHqbWdOSEDs6h7CyyZbxWYTNS+QDYamXgdojJSAceU6nu+1di+APhvHYnsyazL9+fBg7jg3hU5cuK5gssmzJp4HVLgiXSQZQrI2t1XktCREOC9z19BE02M34pMoVXjOhRRbQa0IIEkKcLYTYEPn3TL7Hka62eY5oSYi+yCGwXByjzzUtu4NN+YPY/c5Yxnf/gFSbZmmDFbsU7ANIJSDyt/wjW9lkx9YltXho5+m0Ww0KIfDN5zowv8qCv3zXYnUGqAH5NLDazeEzyQCKtXFhDXrGpvPWO+O3B3pxoNuFz1+5WpON/lS4FIRCbc12jE0F4PT4ZmYABZYFBEQygZwTCGvQA3XnyVH4Q+GsD6JtXVKL3adGk/5x9QZCODk0kVYNoFy67V2L0D06HV3yUurZwwM42OPGne9Zmde8/3xodKjbHD6bDCDZhsiBsHykg076gvjmcx1Y31KFD20szEx3DgAKrY7caR7pH0e/axoOiwHWAozoK5psmPKH0BuZpeTTjs4hmA26aH3/TG1eXAuPN4ijSfoxdw5OICzyuwEc64o189BgN6e1GRwIhfHt549iZZMN15/TouLotNFkt6i6BxDNAGrMPOivnV8Fg45wIA8NYu7bfgKD4z58+f1nad40KhEOAAq1RVINO/o96HVl30hbLdGNYA0OhO04NoStS+uyvrOdKQyXeBlIiwygWCaDDjdvXoiXjzrRPTql6Hse29ONd4Yn8dn3ri6YPPBcUnsGEM0AymIGUGHSo63ZoXqP4O7RKdz/6klct2E+zl1Uo+q1ssEBQKGqSiMWVFegvX+8IM8AyFY2atMestc1jRNDkzk5WdpSU4H5VZakHcLaB8ZRYdSjtVa7rIqbt7ZCR4SHdqbuKzDlD+K//9SJTYtqcHlbYx5Gl39NDgvc0wHVTgNnkwEUa8PCahzsdqW9f5OOrz/bDj0RvnDVatWukQscANKwOpIJ1O8ujEYw8VRVGtHkMOf9LIC8Fn5xDgrRERE2L6nFrndGE564be8fx6oMS0DkSnNVBS5va8Rje7pT/tH7+Z9PYcjjwxeuWl1Q5UNyqSFyGlitDdbOQQ/mV1my3kzd2FqNSX9ItVnymydH8MyhAXzi4mUFuU8YiwNAGtqaHTg5PInRSX/BnQGIJZeEyKcdx4bQXGXJKkMj1pYltRjy+HB65MzlFSEEOgY8mm0Ax7pt22KMTvrx7OH+hF8zNunH/2w/gcvbmrBpcXb7I4VMbg05oNI+wLHBiZwUr9u6tA5EwHOHB3IwqtlCkbTP+VUW3H7R0pw/fq5xAEhDW7MjOm0s5Mi+otGO43nMBAqGwnjt+DAuWtGQs7vbLZE/lPHSQQfGvXBNBTRb/491/vI6LG2w4pdvJN4M/tHLxzHpD+JzV57ZDrOULK23Qq8j3PHIfvzuQG9OX3+5yACSLaiuwEUrGvDorm4EQ6lPnKfjsT3daO8fxxevbkOFqfCzvDgApCH2xGmhbgID0lmA6UAIPWP5yQQ62OOCxxvMaR+C5Y021FpNcfcBtN4AjkVEuHXrIuzrcuFwnNPLPWNT+MUbp3H9OS1pNTApRgtrK/Hw32xFrc2EOx49gA/e9zr2KDjQp0QuMoBi3bK1FQPjXrzY4czJ4wHAuDeA7zx/FJsX1+Cas5tz9rhq4gCQhsV1VliM0v+yQmkEE488Tc7XPsArx4ahI+luOFeICJsW1cTNBGqPlOZelWYTGLVcf24LLEYdHtp55izgey90AgT8w3tWajCy/Nu6tA5PfeoC3H3Degy4p/Hh/3kDn3pon+JMqURykQEU67LVjWiusijawFfqhy92YnTKjy9fc1bR7PNwAEiDVBJCuusslLaL8ci/JPnqDbDj2BDObqlGdWVui5ptWVKL0yNTZ+SWt/ePo6Um/yUgEqmqMOIDGxbgt/v74J6eaTLfMTCOJ/b34GPnLS7oGWOu6XSE689twcv/fAnuvHwFXupw4t13v4KvP9uOcW8g9QPEkasMIJlBr8NNm1ux49gQTo9MZv14p0cm8fM/n8IN57ZgXYv2rR6V4gCQpnNaq9FaWwmzoXDX9xwWI5qrLHnZCHZN+fFWj0uVNpTygbK5+wAdA568lYBW6tZtizAdCOGJfT3Rj337uaOwmQ345CWlUfAtXZUmA+68fCVe/udLcO2G+bh/x0lc8u3t+OWbp9Nee89VBlCsj2xeCL2O8PCu7GcBP3zpOPQ6wj9fUVz7PBwA0vS5967Gb/7+PK2HkdKKJjsO97pVzXUGgD8fH0FYABevzH2N8zXNDlhN+lkBQC4Bka8mMEqtXVCFja3V+OWbUrOYXe+M4sUOJ/7+kmU5nxkVm3lVFnznhvV4+tMXYEWjDf/228O46vuv4tVO5WU0cpUBNHdcl7c14td7erJqEnN6ZBJP7u/FLVsXRfsiFAsOAGmqMOmj+c6F7Kq189DpnMC/PnlI1WygHceGYLcY0u7+pYRBr8M5c/YBjg16NC0Bkcxt2xbh5NAkXj8xgm88244mhxkfP2+J1sMqGGsXVOHR27fh/tvORSAUxm0/3YW7nn475RkKOQMokx4Aqdy6bRFGJ/1ZpYTe89JxGHSET1xc+Gmfc3EAKFE3b2nFZy5bjkd3d+Oup99WpU2kEAI7It2/DHp1XkpbFteiY8AD15RUe1/uzby6AAPA1euaUWs14XOPv4V9XS7cefnKokgFzCciwhVnzcNzd16Ej523GD//8yl88N7Xk1awlTOA1MiiOn9ZPRbVVeKhNzNbBjo9Mokn9vfiL7a2Ft3dP8ABoKT943tW4m8vXIIH3jiNbzzbkfMgcNw5gX63V5X1f5ncH2DPKal415F+qQTEIg1LQCRiMepx46aF6HVNY2mDFTecW3oF33LFYtTjq9eehZ99bBMGx714/z2v4aGd8Xstz2wA534GoNMR/mJLK3adGs0oa+6eyNp/sTb24QBQwogI/3J1G27btgg/3nES3/tTZ04f/xW5+5eKAWDDwmqY9LroMlDHgFQColCrK966rRX1NjO+9L421WZFpeSy1U147o4LsXlxLf71ycP4u1/uxdicTmvyH+ZcnTKf64ZNC2HS6/BwmimhXSNT0t3/ltboKehiw6/QEkdEuOvas3Djphb84MVO3Lv9eM4ee0fnMJY1WLFAxRRHi1GPs1uqsDNSF6i931OQ6/+ylppK7PnS5bhsdZPWQykajQ4LHvj4FnzpfW14+agTV35/B14/Phz9vJwBZFcp7bfWasLV6+bhN3t7MOUPKv6+e17uhF5H+PsizvLiAFAGdDrC1z90Nq7bMB/feu4ofvraO1k/pjcQws6TI6re/cu2LKnF4V43Tg5Pwj0dKIgaQCy3dDrC31y4FE9+8nxYzQbc8tOd+OZzHQiEwuh05j4DaK5bti2CxxfE0wf7FH1918gUnthX3Hf/AAeAsqHXEe6+YT2uPGsevvb7I3FPraZj1zuj8AXDWbV/VGrzkloEwyI6RS/kGQDLztoFVfj9Zy7ATZsX4r7tJ/Dh+17Hcac6GUCxNi2qwaomu+KTwT96+Th0Rbz2L+MAUEYMeh1+cPNGXLa6Ef/65GE8vrcn9Tcl8GrnEEx6HbYuVb+65bmLakAE/DrShL1QSkAwdVSaDPj6h87Gfbecg1Mj6mUAxSIi3LKtFW/1uPFWjyvp13aPTuE3+3pw8+aFBV0RQAkOAGXGZNDh3lvOwQXL6/G5xw8qnvLOtePYMDYvqUGlSf22mA6LEWuaHRj3BguqBART11XrmvHsHRfiM5ctx5Xr5ql+vQ9sXIAKoz5lSuiPXj4OHRH+/pLlqo9JbRwAypDFqMf9f3kuNi2qxZ2/OoDn307vEMyA24ujg568LP/I5DaRvPxTXuZXV+CfrliVl6DvsBjxgY3z8dTB2TWdYnWPTuHxvT24aUvx3/0DHADKVqXJgJ99fDPWLajCpx/eN6uGTSo7OtVP/5xra+Q8QBsv/zAV/cUWqabTb/f3xv38vdvlu//iXvuXcQAoYzazAQ/81RZsXFiDf3zsID7764OK0uB2HBtCo92M1Xn8Y/yuZXVYPc+OS1eXZj9dVhjWtVRhfUsVHnzzzENp3aNT+PUe6e6/kBtCpYMDQJmrqjDi4b/dis9cthyP7+vBtff8GR0D4wm/PhQWeO34MC7MYfcvJaorTXjuzouwsbUmb9dk5emWrYvQ6ZzA7sjpc9m920+U1N0/wAGAQcoO+qcrVuHBv94K11QA193zZzyyqyvusfxDvW64pgK4SIXqn4wVgvevnw+7xTArVbpnbAq/3tONj2wunbt/gAMAi3H+8no8e8eF2LKkFl984hA+88h+eOY08NhxbAhEwIV53ABmLJ8qTHpcf04Lnj00gJEJHwDp7p8IJXX3D3AAYHM02M144ONb8Nn3rsKzhwfwvh+8Nisv+tXOIaxbUIVaa3nXuGel7ZatrfCHwvj13h70uqajd/+l1tmNAwA7g05H+NSly/Gr27chGArj+vtex09fewfu6QD2dblw4Qpe/mGlbUWTHVuX1OLhnV245yWpflYp5P3PxQGAJbRpcS2eueNCXLyyEV/7/RFcf9/rCIVFXvP/GdPKLdsWoWt0Co/s6sKNmxaqWvRQK5oEACK6koiOEtFxIvqCFmNgylRXmvC/f3kuvnzNGpwemYTNbMA5izgTh5W+957VhDqrCUY94ZOXlt7dPwCof45/DiLSA/gRgPcA6AGwm4ieEkIcyfdYmDJEhL+6YAnOX16PCV8ARq5zz8qA2aDHf3xgLTzeYEne/QMaBAAAWwAcF0KcBAAiehTAdQA4ABQ4LsLGys1V65q1HoKqtLiVWwCgO+b9nsjHZiGi24loDxHtGRoaytvgGGOsXGgRAOIdHz3jxJEQ4n4hxCYhxKaGBt50ZIyxXNMiAPQAWBjzfguAzGoSM8YYy5gWAWA3gBVEtISITABuAvCUBuNgjLGylvdNYCFEkIg+DeB5AHoAPxNCvJ3vcTDGWLnTIgsIQohnADyjxbUZY4xJOKGbMcbKFAcAxhgrUxSv5nuhIaIhAKdTfmF89QCGczicQlBqz6nUng9Qes+p1J4PUHrPKd7zWSSESJhHXxQBIBtEtEcIsUnrceRSqT2nUns+QOk9p1J7PkDpPadMng8vATHGWJniAMAYY2WqHALA/VoPQAWl9pxK7fkApfecSu35AKX3nNJ+PiW/B8AYYyy+cpgBMMYYi4MDAGOMlamSDgCl1nqSiE4R0SEiOkBEe7QeTyaI6GdE5CSiwzEfqyWiF4ioM/Lfouk5meD5fJWIeiM/pwNEdLWWY0wHES0kopeJqJ2I3iaiOyIfL+afUaLnVJQ/JyKyENEuIjoYeT53RT6e9s+oZPcAIq0njyGm9SSAm4u59SQRnQKwSQhRtIdXiOgiABMAfiGEWBv52LcAjAohvhEJ1DVCiM9rOU6lEjyfrwKYEEJ8R8uxZYKImgE0CyH2EZEdwF4AHwDwMRTvzyjRc7oRRfhzIiICYBVCTBCREcBrAO4A8CGk+TMq5RlAtPWkEMIPQG49yTQkhNgBYHTOh68D8EDk7Qcg/XIWhQTPp2gJIfqFEPsib3sAtEPq2FfMP6NEz6koCclE5F1j5J9ABj+jUg4AilpPFhkB4I9EtJeIbtd6MDnUJIToB6RfVgCNGo8nFz5NRG9FloiKZrkkFhEtBrARwE6UyM9oznMCivTnRER6IjoAwAngBSFERj+jUg4AilpPFpnzhRDnALgKwKciyw+s8NwHYBmADQD6Adyt6WgyQEQ2AL8BcKcQYlzr8eRCnOdUtD8nIURICLEBUkfFLUS0NpPHKeUAUHKtJ4UQfZH/OgE8CWmZqxQMRtZp5fVap8bjyYoQYjDyCxoG8L8osp9TZF35NwAeEkI8EflwUf+M4j2nYv85AYAQwgVgO4ArkcHPqJQDQEm1niQia2QDC0RkBXAFgMPJv6toPAXgo5G3PwrgdxqOJWvyL2HEB1FEP6fIBuNPAbQLIb4b86mi/Rklek7F+nMiogYiqo68XQHgcgAdyOBnVLJZQAAQSev6b8y0nvxPbUeUOSJaCumuH5A6uT1cjM+HiB4BcAmk0rWDAL4C4LcAHgPQCqALwA1CiKLYWE3wfC6BtKwgAJwC8Hfy2myhI6ILALwK4BCAcOTD/wJpzbxYf0aJntPNKMKfExGdDWmTVw/pJv4xIcS/E1Ed0vwZlXQAYIwxllgpLwExxhhLggMAY4yVKQ4AjDFWpjgAMMZYmeIAwBhjZYoDAGMqI6JLiOj3Wo+Dsbk4ADDGWJniAMBYBBHdGqmzfoCIfhwpuDVBRHcT0T4iepGIGiJfu4GI3owUEntSLiRGRMuJ6E+RWu37iGhZ5OFtRPQ4EXUQ0UOR06mMaYoDAGMAiKgNwEcgFdzbACAE4BYAVgD7IkX4XoF00hcAfgHg80KIsyGdMJU//hCAHwkh1gM4D1KRMUCqQHkngDUAlgI4X+WnxFhKBq0HwFiBeDeAcwHsjtycV0AqphUG8KvI1zwI4AkiqgJQLYR4JfLxBwD8OlKraYEQ4kkAEEJ4ASDyeLuEED2R9w8AWAypkQdjmuEAwJiEADwghPjirA8S/ducr0tWOyXZso4v5u0Q+HePFQBeAmJM8iKADxNRIxDtr7oI0u/IhyNf8xcAXhNCuAGMEdGFkY/fBuCVSI35HiL6QOQxzERUmc8nwVg6+C6EMQBCiCNE9CVIHdd0AAIAPgVgEsBZRLQXgBvSPgEgldv9n8gf+JMAPh75+G0AfkxE/x55jBvy+DQYSwtXA2UsCSKaEELYtB4HY2rgJSDGGCtTPANgjLEyxTMAxhgrUxwAGGOsTHEAYIyxMsUBgDHGyhQHAMYYK1P/Hy1sYbjVzdmeAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['BPPV (HC) R']] [['BPPV (HC) L']]\n",
      "[['BPPV (HC) R']] [['Stroke']]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']]\n",
      "16 3 0.8421052631578947\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3601b39c950548e0bc64e9aa6f4fb1b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/30 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n",
      "1\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[['BPPV (HC) R']] [['BPPV (HC) L']]\n",
      "18 1 0.9473684210526315\n",
      "0.9263157894736842\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 432x288 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "%run model_training.py"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Feature Importance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "with open('encoding_col_map.pkl', 'rb') as f:\n",
    "    col_map = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'Sex': {'indx': [0], 'columns': ['Sex']},\n",
       " 'Stoke Risk (ABCD2 > 4)': {'indx': [1],\n",
       "  'columns': ['Stoke Risk (ABCD2 > 4)']},\n",
       " 'Dizziness Type': {'indx': [2], 'columns': ['Dizziness Type']},\n",
       " 'Triggered by turning in bed?': {'indx': [3],\n",
       "  'columns': ['Triggered by turning in bed?']},\n",
       " 'SG Vertical N': {'indx': [4, 5],\n",
       "  'columns': ['SG Vertical N-None',\n",
       "   'SG Vertical N-Down-beating_vs_Up-beating']},\n",
       " 'SG Horizintal N': {'indx': [6, 7],\n",
       "  'columns': ['SG Horizintal N-None',\n",
       "   'SG Horizintal N-Left-beating_vs_Right-beating']},\n",
       " 'SG Torsional N': {'indx': [8, 9],\n",
       "  'columns': ['SG Torsional N-None',\n",
       "   'SG Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Right Gaze Vertical N': {'indx': [10], 'columns': ['Right Gaze Vertical N']},\n",
       " 'Right G Horizintal N': {'indx': [11], 'columns': ['Right G Horizintal N']},\n",
       " 'Right G Torsional N': {'indx': [12, 13],\n",
       "  'columns': ['Right G Torsional N-None',\n",
       "   'Right G Torsional N-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Left Gaze \"Induced\" Vertical Nystagmus': {'indx': [14],\n",
       "  'columns': ['Left Gaze \"Induced\" Vertical Nystagmus']},\n",
       " 'Left Gaze \"Induced\" Horizintal Nystagmus': {'indx': [15],\n",
       "  'columns': ['Left Gaze \"Induced\" Horizintal Nystagmus']},\n",
       " 'Left Gaze \"Induced\" Torsional Nystagmus': {'indx': [16, 17],\n",
       "  'columns': ['Left Gaze \"Induced\" Torsional Nystagmus-None',\n",
       "   'Left Gaze \"Induced\" Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Up Gaze \"Induced\" Vertical Nystagmus': {'indx': [18],\n",
       "  'columns': ['Up Gaze \"Induced\" Vertical Nystagmus']},\n",
       " 'Up Gaze \"Induced\" Horizintal Nystagmus': {'indx': [19, 20],\n",
       "  'columns': ['Up Gaze \"Induced\" Horizintal Nystagmus-None',\n",
       "   'Up Gaze \"Induced\" Horizintal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Up Gaze \"Induced\" Torsional Nystagmus': {'indx': [21, 22],\n",
       "  'columns': ['Up Gaze \"Induced\" Torsional Nystagmus-None',\n",
       "   'Up Gaze \"Induced\" Torsional Nystagmus-Top pole to right ear torsional nystamgus_vs_Top-pole-to-L-ear torsional']},\n",
       " 'Down Gaze \"Induced\" Vertical Nystagmus': {'indx': [23],\n",
       "  'columns': ['Down Gaze \"Induced\" Vertical Nystagmus']},\n",
       " 'Down Gaze \"Induced\" Horizintal Nystagmus': {'indx': [24, 25],\n",
       "  'columns': ['Down Gaze \"Induced\" Horizintal Nystagmus-None',\n",
       "   'Down Gaze \"Induced\" Horizintal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Down Gaze \"Induced\" Torsional Nystagmus': {'indx': [26],\n",
       "  'columns': ['Down Gaze \"Induced\" Torsional Nystagmus']},\n",
       " 'Right Dix-Hallpike Induced Vertical Nystagmus': {'indx': [27, 28],\n",
       "  'columns': ['Right Dix-Hallpike Induced Vertical Nystagmus-None',\n",
       "   'Right Dix-Hallpike Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Right Dix-Hallpike Induced Horizontal Nystagmus': {'indx': [29, 30],\n",
       "  'columns': ['Right Dix-Hallpike Induced Horizontal Nystagmus-None',\n",
       "   'Right Dix-Hallpike Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Right Dix-Hallpike Induced Torsional Nystagmus': {'indx': [31, 32],\n",
       "  'columns': ['Right Dix-Hallpike Induced Torsional Nystagmus-None',\n",
       "   'Right Dix-Hallpike Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Left Dix-Hallpike Induced Vertical Nystagmus': {'indx': [33, 34],\n",
       "  'columns': ['Left Dix-Hallpike Induced Vertical Nystagmus-None',\n",
       "   'Left Dix-Hallpike Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Left Dix-Hallpike Induced Horizontal Nystagmus': {'indx': [35, 36],\n",
       "  'columns': ['Left Dix-Hallpike Induced Horizontal Nystagmus-None',\n",
       "   'Left Dix-Hallpike Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Left Dix-Hallpike Induced Torsional Nystagmus': {'indx': [37, 38],\n",
       "  'columns': ['Left Dix-Hallpike Induced Torsional Nystagmus-None',\n",
       "   'Left Dix-Hallpike Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Right Roll Induced Vertical Nystagmus': {'indx': [39, 40],\n",
       "  'columns': ['Right Roll Induced Vertical Nystagmus-None',\n",
       "   'Right Roll Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Right Roll Induced Horizontal Nystagmus': {'indx': [41, 42],\n",
       "  'columns': ['Right Roll Induced Horizontal Nystagmus-None',\n",
       "   'Right Roll Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Right Roll Induced Torsional Nystagmus': {'indx': [43, 44],\n",
       "  'columns': ['Right Roll Induced Torsional Nystagmus-None',\n",
       "   'Right Roll Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Left Roll Induced Vertical Nystagmus': {'indx': [45, 46],\n",
       "  'columns': ['Left Roll Induced Vertical Nystagmus-None',\n",
       "   'Left Roll Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Left Roll Induced Horizontal Nystagmus': {'indx': [47, 48],\n",
       "  'columns': ['Left Roll Induced Horizontal Nystagmus-None',\n",
       "   'Left Roll Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Left Roll Induced Torsional Nystagmus': {'indx': [49, 50],\n",
       "  'columns': ['Left Roll Induced Torsional Nystagmus-None',\n",
       "   'Left Roll Induced Torsional Nystagmus-Top-pole-to-L-ear torsional_vs_Top-pole-to-R-ear torsional']},\n",
       " 'Bow Induced Vertical Nystagmus': {'indx': [51, 52],\n",
       "  'columns': ['Bow Induced Vertical Nystagmus-None',\n",
       "   'Bow Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Bow Induced Horizontal Nystagmus': {'indx': [53, 54],\n",
       "  'columns': ['Bow Induced Horizontal Nystagmus-None',\n",
       "   'Bow Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Bow Induced Torsional Nystagmus': {'indx': [55, 56],\n",
       "  'columns': ['Bow Induced Torsional Nystagmus-None',\n",
       "   'Bow Induced Torsional Nystagmus-Torsional Top-Pole-to-the-Left-Ear_vs_Torsional Top-Pole-to-the-Right-Ear']},\n",
       " 'Lean Induced Vertical Nystagmus': {'indx': [57, 58],\n",
       "  'columns': ['Lean Induced Vertical Nystagmus-None',\n",
       "   'Lean Induced Vertical Nystagmus-Down-beating_vs_Up-beating']},\n",
       " 'Lean Induced Horizontal Nystagmus': {'indx': [59, 60],\n",
       "  'columns': ['Lean Induced Horizontal Nystagmus-None',\n",
       "   'Lean Induced Horizontal Nystagmus-Left-beating_vs_Right-beating']},\n",
       " 'Lean Induced Torsional Nystagmus': {'indx': [61, 62],\n",
       "  'columns': ['Lean Induced Torsional Nystagmus-None',\n",
       "   'Lean Induced Torsional Nystagmus-Torsional Top-Pole-to-the-Left-Ear_vs_Torsional Top-Pole-to-the-Right-Ear']},\n",
       " 'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused vertical': {'indx': [63,\n",
       "   64],\n",
       "  'columns': ['Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused vertical-No eye motion',\n",
       "   'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused vertical-Left eye moves down (L hyper)_vs_Left eye moves up (R hyper)']},\n",
       " 'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal': {'indx': [65,\n",
       "   66],\n",
       "  'columns': ['Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-No eye motion',\n",
       "   'Alternate Cover (Moving from Left Eye Occluded to Right Eye Occluded) caused horizontal-L eye moves up (L hyper)_vs_R eye moves down (R hyper)']},\n",
       " 'Head impulse test to the left showed': {'indx': [67],\n",
       "  'columns': ['Head impulse test to the left showed']},\n",
       " 'Head impulse test to the right showed': {'indx': [68],\n",
       "  'columns': ['Head impulse test to the right showed']},\n",
       " 'Gait': {'indx': [69], 'columns': ['Gait']},\n",
       " 'Finger-nose-finger left hand': {'indx': [70, 71],\n",
       "  'columns': ['Finger-nose-finger left hand-Abnormal',\n",
       "   'Finger-nose-finger left hand-None_vs_Normal']},\n",
       " 'Finger-nose-finger right hand': {'indx': [72, 73],\n",
       "  'columns': ['Finger-nose-finger right hand-Abnormal',\n",
       "   'Finger-nose-finger right hand-None_vs_Normal']},\n",
       " 'Hearing loss right ear': {'indx': [74],\n",
       "  'columns': ['Hearing loss right ear']},\n",
       " 'Hearing loss left ear': {'indx': [75], 'columns': ['Hearing loss left ear']}}"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "col_map"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "metadata": {},
   "outputs": [],
   "source": [
    "def relative_feature_importance(values):\n",
    "    r = 20\n",
    "    rescale_max = 0.2\n",
    "    values = 2*np.arccos(values)/np.pi\n",
    "    print(values)\n",
    "    print(np.max(values), np.min(values))\n",
    "    center = ((np.max(values) + np.min(values))/2)\n",
    "    print(center)\n",
    "    values = values - center\n",
    "    print(values)\n",
    "    low = np.min(values)\n",
    "    high = np.max(values)\n",
    "    print(high, low, high-low)\n",
    "    values = (values / max(rescale_max, (high-low))) + .5\n",
    "    print(np.max(values), np.min(values))\n",
    "    print(values)\n",
    "    print((r**(values) - 1)/(r-1))\n",
    "    return (r**(values) - 1)/(r-1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 101,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[0.65481564 0.61602381 0.68344617 0.7476419  0.61602381 0.73210737\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.62026441 0.69659803 0.61602381 0.61602381 0.76187554\n",
      " 0.69659803 0.61602381 0.8134058  0.61602381 0.61602381 0.5666463\n",
      " 0.61602381 0.61602381 0.74905274 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.61602381 0.69782755 0.61602381 0.61602381 0.65481564\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.58023878 0.57905345\n",
      " 0.5666463  0.53089457 0.53089457 0.53089457]\n",
      "0.8134057988020574 0.5308945652761518\n",
      "0.6721501820391046\n",
      "[-0.01733455 -0.05612637  0.01129599  0.07549172 -0.05612637  0.05995719\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05188577  0.02444785 -0.05612637 -0.05612637  0.08972536\n",
      "  0.02444785 -0.05612637  0.14125562 -0.05612637 -0.05612637 -0.10550388\n",
      " -0.05612637 -0.05612637  0.07690256 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05612637  0.02567737 -0.05612637 -0.05612637 -0.01733455\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.0919114  -0.09309674\n",
      " -0.10550388 -0.14125562 -0.14125562 -0.14125562]\n",
      "0.14125561676295284 -0.14125561676295284 0.2825112335259057\n",
      "1.0 0.0\n",
      "[0.43864121 0.30133047 0.5399842  0.7672167  0.30133047 0.7122294\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.31634085 0.5865376  0.30133047 0.30133047 0.81759925\n",
      " 0.5865376  0.30133047 1.         0.30133047 0.30133047 0.12654979\n",
      " 0.30133047 0.30133047 0.77221063 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.30133047 0.59088972 0.30133047 0.30133047 0.43864121\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.17466285 0.17046714\n",
      " 0.12654979 0.         0.         0.        ]\n",
      "[0.14322212 0.07717191 0.21269578 0.47147346 0.07717191 0.39187502\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.08314202 0.25240324 0.07717191 0.07717191 0.55685977\n",
      " 0.25240324 0.07717191 1.         0.07717191 0.07717191 0.02426225\n",
      " 0.07717191 0.07717191 0.47937327 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.07717191 0.25640625 0.07717191 0.07717191 0.14322212\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.03618377 0.03507441\n",
      " 0.02426225 0.         0.         0.        ]\n",
      "Sex                        \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Dizziness Type             \t Vestibular Neuritis L \t 0.477 \t 0.2127 \t [[0.1  0.05 0.01 0.03 0.19 0.6  0.02]]\n",
      "Triggered by turning in bed? \t Vestibular Neuritis L \t 0.3861 \t 0.47147 \t [[0.08 0.04 0.01 0.03 0.37 0.44 0.03]]\n",
      "SG Vertical N              \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "SG Horizintal N            \t Vestibular Neuritis L \t 0.4085 \t 0.39188 \t [[0.14 0.07 0.02 0.04 0.23 0.5  0.02]]\n",
      "SG Torsional N             \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Gaze Vertical N      \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right G Horizintal N       \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right G Torsional N        \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Up Gaze \"Induced\" Vertical N \t Vestibular Neuritis L \t 0.56174 \t 0.08314 \t [[0.04 0.02 0.01 0.02 0.18 0.72 0.02]]\n",
      "Up Gaze \"Induced\" Horizintal \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.04 0.02 0.01 0.01 0.36 0.54 0.02]]\n",
      "Up Gaze \"Induced\" Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Down Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Down Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.36538 \t 0.55686 \t [[0.13 0.07 0.02 0.02 0.29 0.42 0.05]]\n",
      "Down Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.04 0.02 0.01 0.01 0.36 0.54 0.02]]\n",
      "Right Dix-Hallpike Induced V \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.28892 \t 1.0 \t [[0.06 0.04 0.01 0.02 0.56 0.27 0.05]]\n",
      "Right Dix-Hallpike Induced T \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Dix-Hallpike Induced Ve \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Dix-Hallpike Induced Ho \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.04 0.02 0.01 0.01 0.07 0.84 0.02]]\n",
      "Left Dix-Hallpike Induced To \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Roll Induced Vertical  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Roll Induced Horizonta \t Vestibular Neuritis L \t 0.38406 \t 0.47937 \t [[0.08 0.04 0.01 0.03 0.37 0.44 0.03]]\n",
      "Right Roll Induced Torsional \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Vertical N \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Horizontal \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Bow Induced Vertical Nystagm \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Bow Induced Horizontal Nysta \t Vestibular Neuritis L \t 0.45703 \t 0.25641 \t [[0.04 0.02 0.01 0.02 0.36 0.54 0.02]]\n",
      "Bow Induced Torsional Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Lean Induced Vertical Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Lean Induced Horizontal Nyst \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Lean Induced Torsional Nysta \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Head impulse test to the lef \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Head impulse test to the rig \t Vestibular Neuritis L \t 0.61261 \t 0.03618 \t [[0.03 0.02 0.01 0.01 0.12 0.81 0.01]]\n",
      "Gait                       \t Vestibular Neuritis L \t 0.61408 \t 0.03507 \t [[0.02 0.01 0.   0.01 0.15 0.8  0.01]]\n",
      "Finger-nose-finger left hand \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.04 0.02 0.01 0.01 0.07 0.84 0.02]]\n",
      "Finger-nose-finger right han \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss right ear     \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss left ear      \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n"
     ]
    }
   ],
   "source": [
    "indx = 5\n",
    "for _ in range(5):\n",
    "    x = np.array([bm.sample_hidden(X_test[indx])])\n",
    "    baseline = clf.predict_proba(x)\n",
    "    print(np.round(baseline, 2))\n",
    "    pred_indx = clf.predict(x)\n",
    "    y_est = pred_indx.reshape((1, 1))\n",
    "    print(encoder.inverse_transform(y_test[indx].reshape(1, -1)), encoder.inverse_transform(y_est), pred_indx)\n",
    "\n",
    "y_ests = []\n",
    "d_prods = []\n",
    "probs = []\n",
    "for col in col_map:\n",
    "    mask = [False] * X_train.shape[1]\n",
    "    for i in col_map[col]['indx']:\n",
    "        mask[i] = True\n",
    "    hidden = np.array([bm.sample_hidden(X_test[indx], mask=mask, use_visible=False)])\n",
    "    masked_probs = clf.predict_proba(hidden)\n",
    "    pred_indx = clf.predict(hidden)\n",
    "    y_ests.append(pred_indx.reshape((1, 1)))\n",
    "    d_prods.append(np.dot(baseline, masked_probs.T)[0, 0])\n",
    "    probs.append(masked_probs)\n",
    "\n",
    "d_prods2 = relative_feature_importance(np.array(d_prods))\n",
    "\n",
    "for col, y_est, d_prod, d2, masked_probs in zip(col_map, y_ests, d_prods, d_prods2, probs):\n",
    "    print('{:26s} \\t {:20s} \\t'.format(col[:28], encoder.inverse_transform(y_est)[0, 0]), np.round(d_prod, 5), '\\t', np.round(d2, 5), '\\t', np.round(masked_probs, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 98,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[0.65481564 0.61602381 0.68344617 0.7476419  0.61602381 0.73210737\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.62026441 0.69659803 0.61602381 0.61602381 0.76187554\n",
      " 0.69659803 0.61602381 0.8134058  0.61602381 0.61602381 0.5666463\n",
      " 0.61602381 0.61602381 0.74905274 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.61602381 0.69782755 0.61602381 0.61602381 0.65481564\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.58023878 0.57905345\n",
      " 0.5666463  0.53089457 0.53089457 0.53089457]\n",
      "0.8134057988020574 0.5308945652761518\n",
      "0.6721501820391046\n",
      "[-0.01733455 -0.05612637  0.01129599  0.07549172 -0.05612637  0.05995719\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05188577  0.02444785 -0.05612637 -0.05612637  0.08972536\n",
      "  0.02444785 -0.05612637  0.14125562 -0.05612637 -0.05612637 -0.10550388\n",
      " -0.05612637 -0.05612637  0.07690256 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05612637  0.02567737 -0.05612637 -0.05612637 -0.01733455\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.0919114  -0.09309674\n",
      " -0.10550388 -0.14125562 -0.14125562 -0.14125562]\n",
      "0.14125561676295284 -0.14125561676295284 0.2825112335259057\n",
      "1.0 0.0\n",
      "[0.43864121 0.30133047 0.5399842  0.7672167  0.30133047 0.7122294\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.31634085 0.5865376  0.30133047 0.30133047 0.81759925\n",
      " 0.5865376  0.30133047 1.         0.30133047 0.30133047 0.12654979\n",
      " 0.30133047 0.30133047 0.77221063 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.30133047 0.59088972 0.30133047 0.30133047 0.43864121\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.17466285 0.17046714\n",
      " 0.12654979 0.         0.         0.        ]\n",
      "[0.14322212 0.07717191 0.21269578 0.47147346 0.07717191 0.39187502\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.08314202 0.25240324 0.07717191 0.07717191 0.55685977\n",
      " 0.25240324 0.07717191 1.         0.07717191 0.07717191 0.02426225\n",
      " 0.07717191 0.07717191 0.47937327 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.07717191 0.25640625 0.07717191 0.07717191 0.14322212\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.03618377 0.03507441\n",
      " 0.02426225 0.         0.         0.        ]\n",
      "Sex                        \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Dizziness Type             \t Vestibular Neuritis L \t 0.477 \t 0.2127 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Triggered by turning in bed? \t Vestibular Neuritis L \t 0.3861 \t 0.47147 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "SG Vertical N              \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "SG Horizintal N            \t Vestibular Neuritis L \t 0.4085 \t 0.39188 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "SG Torsional N             \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Gaze Vertical N      \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right G Horizintal N       \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right G Torsional N        \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Up Gaze \"Induced\" Vertical N \t Vestibular Neuritis L \t 0.56174 \t 0.08314 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Up Gaze \"Induced\" Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Down Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.36538 \t 0.55686 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Down Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Dix-Hallpike Induced V \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.28892 \t 1.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Dix-Hallpike Induced T \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Dix-Hallpike Induced Ho \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Dix-Hallpike Induced To \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Roll Induced Vertical  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Roll Induced Horizonta \t Vestibular Neuritis L \t 0.38406 \t 0.47937 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Right Roll Induced Torsional \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Roll Induced Vertical N \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Roll Induced Horizontal \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Roll Induced Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Bow Induced Vertical Nystagm \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Bow Induced Horizontal Nysta \t Vestibular Neuritis L \t 0.45703 \t 0.25641 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Bow Induced Torsional Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Lean Induced Vertical Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Lean Induced Horizontal Nyst \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Lean Induced Torsional Nysta \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Head impulse test to the lef \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Head impulse test to the rig \t Vestibular Neuritis L \t 0.61261 \t 0.03618 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Gait                       \t Vestibular Neuritis L \t 0.61408 \t 0.03507 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Finger-nose-finger left hand \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Finger-nose-finger right han \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss right ear     \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss left ear      \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n"
     ]
    }
   ],
   "source": [
    "# indx = 0\n",
    "for _ in range(5):\n",
    "    x = np.array([bm.sample_hidden(X_test[indx])])\n",
    "    baseline = clf.predict_proba(x)\n",
    "    print(np.round(baseline, 2))\n",
    "    pred_indx = clf.predict(x)\n",
    "    y_est = pred_indx.reshape((1, 1))\n",
    "    print(encoder.inverse_transform(y_test[indx].reshape(1, -1)), encoder.inverse_transform(y_est), pred_indx)\n",
    "\n",
    "y_ests = []\n",
    "d_prods = []\n",
    "probs = []\n",
    "for col in col_map:\n",
    "    mask = [False] * X_train.shape[1]\n",
    "    for i in col_map[col]['indx']:\n",
    "        mask[i] = True\n",
    "    hidden = np.array([bm.sample_hidden(X_test[indx], mask=mask, use_visible=False)])\n",
    "    masked_probs = clf.predict_proba(hidden)\n",
    "    pred_indx = clf.predict(hidden)\n",
    "    y_ests.append(pred_indx.reshape((1, 1)))\n",
    "    d_prods.append(np.dot(baseline, masked_probs.T)[0, 0])\n",
    "    probs.append(masked_probs)\n",
    "\n",
    "d_prods2 = relative_feature_importance(np.array(d_prods))\n",
    "\n",
    "for col, y_est, d_prod, d2, masked_probs in zip(col_map, y_ests, d_prods, d_prods2, probs):\n",
    "    print('{:26s} \\t {:20s} \\t'.format(col[:28], encoder.inverse_transform(y_est)[0, 0]), np.round(d_prod, 5), '\\t', np.round(d2, 5), '\\t', np.round(masked_probs, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 99,
   "metadata": {
    "scrolled": false
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[0.25869898 0.25869898 0.45714238 0.25370389 0.2635879  0.25869898\n",
      " 0.30915824 0.23032457 0.2262684  0.21956069 0.25430677 0.25869898\n",
      " 0.24806327 0.25869898 0.25869898 0.35167572 0.2702719  0.25869898\n",
      " 0.25869898 0.27584499 0.25869898 0.27975113 0.24444762 0.25869898\n",
      " 0.25579236 0.26447827 0.25869898 0.28533775 0.26447827 0.25869898\n",
      " 0.39328062 0.2635399  0.25869898 0.2174339  0.22593596 0.25869898\n",
      " 0.26670749 0.27584499 0.26899156 0.25869898 0.25869898 0.25869898\n",
      " 0.22597834 0.23615868 0.23032457 0.24957594]\n",
      "0.45714238244894057 0.2174339014940353\n",
      "0.33728814197148793\n",
      "[-0.07858916 -0.07858916  0.11985424 -0.08358425 -0.07370024 -0.07858916\n",
      " -0.0281299  -0.10696358 -0.11101974 -0.11772745 -0.08298137 -0.07858916\n",
      " -0.08922487 -0.07858916 -0.07858916  0.01438758 -0.06701624 -0.07858916\n",
      " -0.07858916 -0.06144316 -0.07858916 -0.05753701 -0.09284052 -0.07858916\n",
      " -0.08149578 -0.07280987 -0.07858916 -0.05195039 -0.07280987 -0.07858916\n",
      "  0.05599248 -0.07374824 -0.07858916 -0.11985424 -0.11135218 -0.07858916\n",
      " -0.07058065 -0.06144316 -0.06829658 -0.07858916 -0.07858916 -0.07858916\n",
      " -0.1113098  -0.10112946 -0.10696358 -0.0877122 ]\n",
      "0.11985424047745263 -0.11985424047745263 0.23970848095490527\n",
      "1.0 0.0\n",
      "[0.17214692 0.17214692 1.         0.15130876 0.19254219 0.17214692\n",
      " 0.38264952 0.05377642 0.03685519 0.0088724  0.15382378 0.17214692\n",
      " 0.12777756 0.17214692 0.17214692 0.56002116 0.22042609 0.17214692\n",
      " 0.17214692 0.2436755  0.17214692 0.25997089 0.11269406 0.17214692\n",
      " 0.16002127 0.19625659 0.17214692 0.28327679 0.19625659 0.17214692\n",
      " 0.73358573 0.19234197 0.17214692 0.         0.03546832 0.17214692\n",
      " 0.20555629 0.2436755  0.21508483 0.17214692 0.17214692 0.17214692\n",
      " 0.03564511 0.0781148  0.05377642 0.13408803]\n",
      "[0.03551688 0.03551688 1.         0.0301824  0.04107057 0.03551688\n",
      " 0.11297779 0.00920011 0.00614389 0.00141767 0.0308087  0.03551688\n",
      " 0.02454559 0.03551688 0.03551688 0.22910986 0.04923397 0.03551688\n",
      " 0.03551688 0.05658172 0.03551688 0.06204542 0.02113587 0.03551688\n",
      " 0.03237233 0.04211905 0.03551688 0.07033809 0.04211905 0.03551688\n",
      " 0.42124303 0.04101439 0.03551688 0.         0.0059002  0.03551688\n",
      " 0.04479586 0.05658172 0.047617   0.03551688 0.03551688 0.03551688\n",
      " 0.00593121 0.01387676 0.00920011 0.02601846]\n",
      "Sex                        \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Dizziness Type             \t Stroke               \t 0.75307 \t 1.0 \t [[0.   0.02 0.07 0.09 0.78 0.01 0.02]]\n",
      "Triggered by turning in bed? \t Stroke               \t 0.92164 \t 0.03018 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "SG Vertical N              \t Stroke               \t 0.9155 \t 0.04107 \t [[0.   0.   0.02 0.02 0.95 0.   0.  ]]\n",
      "SG Horizintal N            \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "SG Torsional N             \t Stroke               \t 0.88438 \t 0.11298 \t [[0.   0.01 0.04 0.02 0.92 0.   0.01]]\n",
      "Right Gaze Vertical N      \t Stroke               \t 0.93526 \t 0.0092 \t [[0.   0.   0.01 0.   0.98 0.   0.  ]]\n",
      "Right G Horizintal N       \t Stroke               \t 0.9375 \t 0.00614 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right G Torsional N        \t Stroke               \t 0.94111 \t 0.00142 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Vertical \t Stroke               \t 0.92127 \t 0.03081 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Horizint \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Torsiona \t Stroke               \t 0.92504 \t 0.02455 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Vertical N \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Torsional  \t Stroke               \t 0.85126 \t 0.22911 \t [[0.   0.01 0.06 0.03 0.89 0.01 0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t Stroke               \t 0.91123 \t 0.04923 \t [[0.   0.   0.02 0.01 0.95 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Horizint \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Torsiona \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced V \t Stroke               \t 0.90759 \t 0.05658 \t [[0.   0.   0.02 0.02 0.95 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced T \t Stroke               \t 0.90499 \t 0.06205 \t [[0.   0.01 0.02 0.01 0.94 0.   0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t Stroke               \t 0.92718 \t 0.02114 \t [[0.   0.   0.01 0.01 0.97 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ho \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced To \t Stroke               \t 0.92036 \t 0.03237 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Right Roll Induced Vertical  \t Stroke               \t 0.91494 \t 0.04212 \t [[0.   0.   0.02 0.01 0.95 0.   0.  ]]\n",
      "Right Roll Induced Horizonta \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Right Roll Induced Torsional \t Stroke               \t 0.90123 \t 0.07034 \t [[0.   0.   0.02 0.02 0.94 0.   0.01]]\n",
      "Left Roll Induced Vertical N \t Stroke               \t 0.91494 \t 0.04212 \t [[0.   0.   0.02 0.01 0.95 0.   0.  ]]\n",
      "Left Roll Induced Horizontal \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Left Roll Induced Torsional  \t Stroke               \t 0.81518 \t 0.42124 \t [[0.01 0.01 0.08 0.02 0.85 0.01 0.02]]\n",
      "Bow Induced Vertical Nystagm \t Stroke               \t 0.91553 \t 0.04101 \t [[0.   0.   0.02 0.02 0.95 0.   0.  ]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Bow Induced Torsional Nystag \t Stroke               \t 0.94224 \t 0.0 \t [[0.   0.   0.01 0.   0.98 0.   0.  ]]\n",
      "Lean Induced Vertical Nystag \t Stroke               \t 0.93768 \t 0.0059 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Lean Induced Torsional Nysta \t Stroke               \t 0.91352 \t 0.0448 \t [[0.   0.   0.02 0.01 0.95 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.90759 \t 0.05658 \t [[0.   0.   0.02 0.02 0.95 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.91205 \t 0.04762 \t [[0.   0.   0.02 0.02 0.95 0.   0.01]]\n",
      "Head impulse test to the lef \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Gait                       \t Stroke               \t 0.91856 \t 0.03552 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Finger-nose-finger left hand \t Stroke               \t 0.93766 \t 0.00593 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Finger-nose-finger right han \t Stroke               \t 0.93198 \t 0.01388 \t [[0.   0.   0.01 0.   0.97 0.   0.  ]]\n",
      "Hearing loss right ear     \t Stroke               \t 0.93526 \t 0.0092 \t [[0.   0.   0.01 0.   0.98 0.   0.  ]]\n",
      "Hearing loss left ear      \t Stroke               \t 0.92413 \t 0.02602 \t [[0.   0.   0.01 0.01 0.96 0.   0.  ]]\n",
      "[[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) R']] [1.]\n",
      "[[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) R']] [1.]\n",
      "[[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) R']] [1.]\n",
      "[[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) R']] [1.]\n",
      "[[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) R']] [1.]\n",
      "[0.81400733 0.73446437 0.73446437 0.73446437 0.73446437 0.73446437\n",
      " 0.73446437 0.73446437 0.77915873 0.73446437 0.73446437 0.73446437\n",
      " 0.73446437 0.75912988 0.73446437 0.73446437 0.73446437 0.73446437\n",
      " 0.74653726 0.73446437 0.73725934 0.73446437 0.73446437 0.78039308\n",
      " 0.73446437 0.73446437 0.73446437 0.73446437 0.73446437 0.72624912\n",
      " 0.73446437 0.73446437 0.82636033 0.73446437 0.73446437 0.77376935\n",
      " 0.73446437 0.73446437 0.73446437 0.73446437 0.73446437 0.73446437\n",
      " 0.77238581 0.83228906 0.76327142 0.73446437]\n",
      "0.8322890634282529 0.7262491153191707\n",
      "0.7792690893737118\n",
      "[ 0.03473824 -0.04480471 -0.04480471 -0.04480471 -0.04480471 -0.04480471\n",
      " -0.04480471 -0.04480471 -0.00011036 -0.04480471 -0.04480471 -0.04480471\n",
      " -0.04480471 -0.02013921 -0.04480471 -0.04480471 -0.04480471 -0.04480471\n",
      " -0.03273183 -0.04480471 -0.04200975 -0.04480471 -0.04480471  0.00112399\n",
      " -0.04480471 -0.04480471 -0.04480471 -0.04480471 -0.04480471 -0.05301997\n",
      " -0.04480471 -0.04480471  0.04709124 -0.04480471 -0.04480471 -0.00549974\n",
      " -0.04480471 -0.04480471 -0.04480471 -0.04480471 -0.04480471 -0.04480471\n",
      " -0.00688327  0.05301997 -0.01599767 -0.04480471]\n",
      "0.053019974054541175 -0.053019974054541064 0.10603994810908224\n",
      "0.7650998702727059 0.23490012972729468\n",
      "[0.67369119 0.27597643 0.27597643 0.27597643 0.27597643 0.27597643\n",
      " 0.27597643 0.27597643 0.49944819 0.27597643 0.27597643 0.27597643\n",
      " 0.27597643 0.39930397 0.27597643 0.27597643 0.27597643 0.27597643\n",
      " 0.33634085 0.27597643 0.28995125 0.27597643 0.27597643 0.50561994\n",
      " 0.27597643 0.27597643 0.27597643 0.27597643 0.27597643 0.23490013\n",
      " 0.27597643 0.27597643 0.7354562  0.27597643 0.27597643 0.47250129\n",
      " 0.27597643 0.27597643 0.27597643 0.27597643 0.27597643 0.27597643\n",
      " 0.46558363 0.76509987 0.42001166 0.27597643]\n",
      "[0.34340824 0.06767795 0.06767795 0.06767795 0.06767795 0.06767795\n",
      " 0.06767795 0.06767795 0.18235523 0.06767795 0.06767795 0.06767795\n",
      " 0.06767795 0.12145001 0.06767795 0.06767795 0.06767795 0.06767795\n",
      " 0.09152549 0.06767795 0.0728216  0.06767795 0.06767795 0.18674029\n",
      " 0.06767795 0.06767795 0.06767795 0.06767795 0.06767795 0.05374806\n",
      " 0.06767795 0.06767795 0.4239058  0.06767795 0.06767795 0.16413122\n",
      " 0.06767795 0.06767795 0.06767795 0.06767795 0.06767795 0.06767795\n",
      " 0.15968537 0.46816039 0.1325911  0.06767795]\n",
      "Sex                        \t BPPV (HC) L          \t 0.28802 \t 0.34341 \t [[0.38 0.31 0.04 0.02 0.18 0.05 0.02]]\n",
      "Stoke Risk (ABCD2 > 4)     \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Dizziness Type             \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Triggered by turning in bed? \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "SG Vertical N              \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "SG Horizintal N            \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "SG Torsional N             \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right Gaze Vertical N      \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right G Horizintal N       \t BPPV (HC) R          \t 0.33998 \t 0.18236 \t [[0.37 0.41 0.02 0.05 0.1  0.03 0.02]]\n",
      "Right G Torsional N        \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Gaze \"Induced\" Vertical \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Gaze \"Induced\" Horizint \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Gaze \"Induced\" Torsiona \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Up Gaze \"Induced\" Vertical N \t BPPV (HC) R          \t 0.3694 \t 0.12145 \t [[0.25 0.51 0.02 0.07 0.1  0.03 0.02]]\n",
      "Up Gaze \"Induced\" Horizintal \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Up Gaze \"Induced\" Torsional  \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Down Gaze \"Induced\" Vertical \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Down Gaze \"Induced\" Horizint \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Down Gaze \"Induced\" Torsiona \t BPPV (HC) R          \t 0.3877 \t 0.09153 \t [[0.2  0.57 0.03 0.04 0.08 0.05 0.03]]\n",
      "Right Dix-Hallpike Induced V \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right Dix-Hallpike Induced H \t BPPV (HC) R          \t 0.4011 \t 0.07282 \t [[0.23 0.58 0.03 0.03 0.05 0.06 0.02]]\n",
      "Right Dix-Hallpike Induced T \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Dix-Hallpike Induced Ve \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Dix-Hallpike Induced Ho \t BPPV (HC) R          \t 0.33816 \t 0.18674 \t [[0.19 0.48 0.02 0.07 0.16 0.06 0.02]]\n",
      "Left Dix-Hallpike Induced To \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right Roll Induced Vertical  \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right Roll Induced Horizonta \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Right Roll Induced Torsional \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Roll Induced Vertical N \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Left Roll Induced Horizontal \t BPPV (HC) R          \t 0.41688 \t 0.05375 \t [[0.27 0.59 0.02 0.03 0.05 0.02 0.01]]\n",
      "Left Roll Induced Torsional  \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Bow Induced Vertical Nystagm \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Bow Induced Horizontal Nysta \t BPPV (HC) L          \t 0.26938 \t 0.42391 \t [[0.64 0.17 0.03 0.02 0.06 0.06 0.03]]\n",
      "Bow Induced Torsional Nystag \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Lean Induced Vertical Nystag \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Lean Induced Horizontal Nyst \t BPPV (HC) R          \t 0.34793 \t 0.16413 \t [[0.36 0.43 0.02 0.03 0.12 0.02 0.02]]\n",
      "Lean Induced Torsional Nysta \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Alternate Cover (Moving from \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Alternate Cover (Moving from \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Head impulse test to the lef \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Head impulse test to the rig \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Gait                       \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "Finger-nose-finger left hand \t BPPV (HC) R          \t 0.34997 \t 0.15969 \t [[0.36 0.43 0.02 0.02 0.14 0.01 0.01]]\n",
      "Finger-nose-finger right han \t BPPV (HC) L          \t 0.2604 \t 0.46816 \t [[0.33 0.28 0.03 0.01 0.31 0.02 0.02]]\n",
      "Hearing loss right ear     \t BPPV (HC) R          \t 0.36334 \t 0.13259 \t [[0.17 0.54 0.03 0.03 0.08 0.13 0.01]]\n",
      "Hearing loss left ear      \t BPPV (HC) R          \t 0.40511 \t 0.06768 \t [[0.26 0.58 0.03 0.03 0.04 0.04 0.02]]\n",
      "[[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']] [0.]\n",
      "[[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']] [0.]\n",
      "[[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']] [0.]\n",
      "[[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']] [0.]\n",
      "[[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "[['BPPV (HC) R']] [['BPPV (HC) L']] [0.]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.80694597 0.81413217 0.80694597 0.80694597 0.80694597 0.80694597\n",
      " 0.80694597 0.80694597 0.81413217 0.80694597 0.80694597 0.80694597\n",
      " 0.80694597 0.79236032 0.80694597 0.80694597 0.80694597 0.80694597\n",
      " 0.83961441 0.80694597 0.80257051 0.80694597 0.80694597 0.80694597\n",
      " 0.80694597 0.80694597 0.79672428 0.80694597 0.80694597 0.80694597\n",
      " 0.80694597 0.80694597 0.80556877 0.80694597 0.80694597 0.79855141\n",
      " 0.80694597 0.80694597 0.80694597 0.80694597 0.80694597 0.80694597\n",
      " 0.80694597 0.80807639 0.81562241 0.80556877]\n",
      "0.8396144090293327 0.7923603203129652\n",
      "0.815987364671149\n",
      "[-0.0090414  -0.00185519 -0.0090414  -0.0090414  -0.0090414  -0.0090414\n",
      " -0.0090414  -0.0090414  -0.00185519 -0.0090414  -0.0090414  -0.0090414\n",
      " -0.0090414  -0.02362704 -0.0090414  -0.0090414  -0.0090414  -0.0090414\n",
      "  0.02362704 -0.0090414  -0.01341685 -0.0090414  -0.0090414  -0.0090414\n",
      " -0.0090414  -0.0090414  -0.01926309 -0.0090414  -0.0090414  -0.0090414\n",
      " -0.0090414  -0.0090414  -0.0104186  -0.0090414  -0.0090414  -0.01743595\n",
      " -0.0090414  -0.0090414  -0.0090414  -0.0090414  -0.0090414  -0.0090414\n",
      " -0.0090414  -0.00791097 -0.00036495 -0.0104186 ]\n",
      "0.023627044358183746 -0.023627044358183746 0.04725408871636749\n",
      "0.6181352217909187 0.38186477820908127\n",
      "[0.45479302 0.49072405 0.45479302 0.45479302 0.45479302 0.45479302\n",
      " 0.45479302 0.45479302 0.49072405 0.45479302 0.45479302 0.45479302\n",
      " 0.45479302 0.38186478 0.45479302 0.45479302 0.45479302 0.45479302\n",
      " 0.61813522 0.45479302 0.43291574 0.45479302 0.45479302 0.45479302\n",
      " 0.45479302 0.45479302 0.40368457 0.45479302 0.45479302 0.45479302\n",
      " 0.45479302 0.45479302 0.44790701 0.45479302 0.45479302 0.41282025\n",
      " 0.45479302 0.45479302 0.45479302 0.45479302 0.45479302 0.45479302\n",
      " 0.45479302 0.46044514 0.49817524 0.44790701]\n",
      "[0.1529318  0.17629336 0.1529318  0.1529318  0.1529318  0.1529318\n",
      " 0.1529318  0.1529318  0.17629336 0.1529318  0.1529318  0.1529318\n",
      " 0.1529318  0.11258892 0.1529318  0.1529318  0.1529318  0.1529318\n",
      " 0.28268796 0.1529318  0.13989148 0.1529318  0.1529318  0.1529318\n",
      " 0.1529318  0.1529318  0.12374956 0.1529318  0.1529318  0.1529318\n",
      " 0.1529318  0.1529318  0.14873475 0.1529318  0.1529318  0.12864343\n",
      " 0.1529318  0.1529318  0.1529318  0.1529318  0.1529318  0.1529318\n",
      " 0.1529318  0.15644209 0.18146083 0.14873475]\n",
      "Sex                        \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t BPPV (HC) L          \t 0.28783 \t 0.17629 \t [[0.38 0.31 0.04 0.02 0.18 0.05 0.02]]\n",
      "Dizziness Type             \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Triggered by turning in bed? \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "SG Vertical N              \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "SG Horizintal N            \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "SG Torsional N             \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Right Gaze Vertical N      \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Right G Horizintal N       \t BPPV (HC) L          \t 0.28783 \t 0.17629 \t [[0.38 0.31 0.04 0.02 0.18 0.05 0.02]]\n",
      "Right G Torsional N        \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Gaze \"Induced\" Vertical \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Gaze \"Induced\" Horizint \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Up Gaze \"Induced\" Vertical N \t BPPV (HC) L          \t 0.32041 \t 0.11259 \t [[0.5  0.31 0.02 0.02 0.12 0.01 0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Up Gaze \"Induced\" Torsional  \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Down Gaze \"Induced\" Horizint \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Down Gaze \"Induced\" Torsiona \t BPPV (HC) L          \t 0.24928 \t 0.28269 \t [[0.42 0.16 0.03 0.02 0.12 0.23 0.02]]\n",
      "Right Dix-Hallpike Induced V \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Right Dix-Hallpike Induced H \t BPPV (HC) R          \t 0.30517 \t 0.13989 \t [[0.36 0.48 0.03 0.02 0.06 0.04 0.01]]\n",
      "Right Dix-Hallpike Induced T \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Dix-Hallpike Induced Ho \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Dix-Hallpike Induced To \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Right Roll Induced Vertical  \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Right Roll Induced Horizonta \t BPPV (HC) R          \t 0.31391 \t 0.12375 \t [[0.4  0.46 0.03 0.02 0.05 0.02 0.01]]\n",
      "Right Roll Induced Torsional \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Roll Induced Vertical N \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Roll Induced Horizontal \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Left Roll Induced Torsional  \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Bow Induced Vertical Nystagm \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Bow Induced Horizontal Nysta \t BPPV (HC) L          \t 0.30069 \t 0.14873 \t [[0.56 0.1  0.03 0.01 0.22 0.06 0.02]]\n",
      "Bow Induced Torsional Nystag \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Lean Induced Vertical Nystag \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Lean Induced Horizontal Nyst \t BPPV (HC) L          \t 0.31118 \t 0.12864 \t [[0.49 0.31 0.03 0.03 0.1  0.02 0.02]]\n",
      "Lean Induced Torsional Nysta \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Alternate Cover (Moving from \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Alternate Cover (Moving from \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Head impulse test to the lef \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Head impulse test to the rig \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Gait                       \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Finger-nose-finger left hand \t BPPV (HC) L          \t 0.29862 \t 0.15293 \t [[0.39 0.31 0.03 0.01 0.22 0.03 0.01]]\n",
      "Finger-nose-finger right han \t BPPV (HC) L          \t 0.29693 \t 0.15644 \t [[0.39 0.31 0.03 0.02 0.21 0.03 0.01]]\n",
      "Hearing loss right ear     \t BPPV (HC) L          \t 0.28559 \t 0.18146 \t [[0.34 0.3  0.03 0.01 0.27 0.05 0.01]]\n",
      "Hearing loss left ear      \t BPPV (HC) L          \t 0.30069 \t 0.14873 \t [[0.56 0.1  0.03 0.01 0.22 0.06 0.02]]\n",
      "[[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "[['BPPV (PC) L']] [['BPPV (PC) L']] [2.]\n",
      "[[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "[['BPPV (PC) L']] [['BPPV (PC) L']] [2.]\n",
      "[[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "[['BPPV (PC) L']] [['BPPV (PC) L']] [2.]\n",
      "[[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "[['BPPV (PC) L']] [['BPPV (PC) L']] [2.]\n",
      "[[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "[['BPPV (PC) L']] [['BPPV (PC) L']] [2.]\n",
      "[0.61896568 0.5711291  0.5711291  0.5711291  0.5711291  0.5711291\n",
      " 0.5711291  0.5711291  0.59680249 0.5711291  0.5711291  0.5711291\n",
      " 0.5711291  0.62649939 0.5711291  0.5711291  0.5711291  0.5711291\n",
      " 0.57929143 0.5711291  0.5711291  0.5711291  0.62413045 0.5711291\n",
      " 0.57601387 0.5711291  0.5711291  0.5711291  0.55995165 0.5711291\n",
      " 0.58154759 0.5711291  0.5711291  0.5711291  0.63229784 0.5711291\n",
      " 0.63073185 0.5711291  0.5711291  0.5711291  0.5711291  0.5711291\n",
      " 0.60045806 0.58499167 0.59998444 0.61257942]\n",
      "0.6322978422983891 0.5599516510276729\n",
      "0.5961247466630311\n",
      "[ 0.02284093 -0.02499565 -0.02499565 -0.02499565 -0.02499565 -0.02499565\n",
      " -0.02499565 -0.02499565  0.00067774 -0.02499565 -0.02499565 -0.02499565\n",
      " -0.02499565  0.03037464 -0.02499565 -0.02499565 -0.02499565 -0.02499565\n",
      " -0.01683332 -0.02499565 -0.02499565 -0.02499565  0.0280057  -0.02499565\n",
      " -0.02011088 -0.02499565 -0.02499565 -0.02499565 -0.0361731  -0.02499565\n",
      " -0.01457716 -0.02499565 -0.02499565 -0.02499565  0.0361731  -0.02499565\n",
      "  0.0346071  -0.02499565 -0.02499565 -0.02499565 -0.02499565 -0.02499565\n",
      "  0.00433332 -0.01113308  0.00385969  0.01645467]\n",
      "0.03617309563535809 -0.0361730956353582 0.07234619127071629\n",
      "0.6808654781767904 0.319134521823209\n",
      "[0.61420467 0.37502175 0.37502175 0.37502175 0.37502175 0.37502175\n",
      " 0.37502175 0.37502175 0.50338871 0.37502175 0.37502175 0.37502175\n",
      " 0.37502175 0.6518732  0.37502175 0.37502175 0.37502175 0.37502175\n",
      " 0.41583341 0.37502175 0.37502175 0.37502175 0.64002852 0.37502175\n",
      " 0.39944559 0.37502175 0.37502175 0.37502175 0.31913452 0.37502175\n",
      " 0.4271142  0.37502175 0.37502175 0.37502175 0.68086548 0.37502175\n",
      " 0.67303551 0.37502175 0.37502175 0.37502175 0.37502175 0.37502175\n",
      " 0.52166658 0.44433461 0.51929846 0.58227335]\n",
      "[0.27876277 0.1092364  0.1092364  0.1092364  0.1092364  0.1092364\n",
      " 0.1092364  0.1092364  0.18514562 0.1092364  0.1092364  0.1092364\n",
      " 0.1092364  0.31835055 0.1092364  0.1092364  0.1092364  0.1092364\n",
      " 0.13028714 0.1092364  0.1092364  0.1092364  0.30541761 0.1092364\n",
      " 0.12152388 0.1092364  0.1092364  0.1092364  0.08428309 0.1092364\n",
      " 0.13657437 0.1092364  0.1092364  0.1092364  0.35201215 0.1092364\n",
      " 0.34263108 0.1092364  0.1092364  0.1092364  0.1092364  0.1092364\n",
      " 0.1985283  0.14659123 0.19675281 0.24853134]\n",
      "Sex                        \t BPPV (PC) L          \t 0.56343 \t 0.27876 \t [[0.01 0.01 0.7  0.1  0.14 0.02 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Dizziness Type             \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Triggered by turning in bed? \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "SG Vertical N              \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "SG Horizintal N            \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "SG Torsional N             \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right Gaze Vertical N      \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right G Horizintal N       \t BPPV (PC) L          \t 0.59184 \t 0.18515 \t [[0.04 0.03 0.75 0.06 0.09 0.02 0.01]]\n",
      "Right G Torsional N        \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Gaze \"Induced\" Vertical \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Gaze \"Induced\" Horizint \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Up Gaze \"Induced\" Vertical N \t BPPV (PC) L          \t 0.55361 \t 0.31835 \t [[0.03 0.02 0.67 0.2  0.05 0.02 0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Up Gaze \"Induced\" Torsional  \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Down Gaze \"Induced\" Horizint \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Down Gaze \"Induced\" Torsiona \t BPPV (PC) L          \t 0.61379 \t 0.13029 \t [[0.03 0.02 0.76 0.13 0.04 0.01 0.  ]]\n",
      "Right Dix-Hallpike Induced V \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right Dix-Hallpike Induced H \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right Dix-Hallpike Induced T \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t BPPV (PC) L          \t 0.55671 \t 0.30542 \t [[0.02 0.03 0.68 0.2  0.05 0.02 0.01]]\n",
      "Left Dix-Hallpike Induced Ho \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Dix-Hallpike Induced To \t BPPV (PC) L          \t 0.61784 \t 0.12152 \t [[0.04 0.03 0.78 0.08 0.05 0.01 0.01]]\n",
      "Right Roll Induced Vertical  \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right Roll Induced Horizonta \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Right Roll Induced Torsional \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Roll Induced Vertical N \t BPPV (PC) L          \t 0.63748 \t 0.08428 \t [[0.02 0.02 0.8  0.08 0.05 0.01 0.01]]\n",
      "Left Roll Induced Horizontal \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Left Roll Induced Torsional  \t BPPV (PC) L          \t 0.61098 \t 0.13657 \t [[0.06 0.04 0.77 0.07 0.03 0.01 0.01]]\n",
      "Bow Induced Vertical Nystagm \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Bow Induced Horizontal Nysta \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Bow Induced Torsional Nystag \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Lean Induced Vertical Nystag \t BPPV (PC) L          \t 0.546 \t 0.35201 \t [[0.05 0.03 0.68 0.1  0.12 0.01 0.01]]\n",
      "Lean Induced Horizontal Nyst \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Lean Induced Torsional Nysta \t BPPV (PC) L          \t 0.54806 \t 0.34263 \t [[0.04 0.03 0.68 0.13 0.09 0.03 0.  ]]\n",
      "Alternate Cover (Moving from \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Alternate Cover (Moving from \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Head impulse test to the lef \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Head impulse test to the rig \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Gait                       \t BPPV (PC) L          \t 0.62386 \t 0.10924 \t [[0.02 0.01 0.78 0.14 0.04 0.01 0.01]]\n",
      "Finger-nose-finger left hand \t BPPV (PC) L          \t 0.5872 \t 0.19853 \t [[0.01 0.01 0.72 0.18 0.05 0.03 0.01]]\n",
      "Finger-nose-finger right han \t BPPV (PC) L          \t 0.60669 \t 0.14659 \t [[0.02 0.02 0.76 0.08 0.09 0.03 0.01]]\n",
      "Hearing loss right ear     \t BPPV (PC) L          \t 0.58781 \t 0.19675 \t [[0.01 0.01 0.74 0.07 0.12 0.04 0.  ]]\n",
      "Hearing loss left ear      \t BPPV (PC) L          \t 0.57169 \t 0.24853 \t [[0.03 0.02 0.7  0.19 0.04 0.02 0.01]]\n",
      "[[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.12126798 0.12126798 0.13164807 0.12126798 0.12915181 0.12126798\n",
      " 0.13192999 0.12446455 0.1241515  0.12915181 0.11324294 0.12126798\n",
      " 0.13298248 0.12126798 0.12126798 0.12126798 0.13611188 0.12126798\n",
      " 0.11324294 0.13452951 0.12126798 0.11324294 0.13919258 0.12126798\n",
      " 0.12697792 0.14945739 0.12126798 0.15932353 0.12301271 0.12126798\n",
      " 0.12027242 0.12126798 0.12126798 0.12126798 0.12126798 0.12126798\n",
      " 0.12126798 0.15391378 0.16173316 0.12126798 0.12126798 0.12126798\n",
      " 0.13452951 0.12126798 0.11585917 0.16173316]\n",
      "0.161733162208161 0.11324293522323886\n",
      "0.13748804871569992\n",
      "[-0.01622007 -0.01622007 -0.00583998 -0.01622007 -0.00833623 -0.01622007\n",
      " -0.00555806 -0.0130235  -0.01333655 -0.00833623 -0.02424511 -0.01622007\n",
      " -0.00450557 -0.01622007 -0.01622007 -0.01622007 -0.00137617 -0.01622007\n",
      " -0.02424511 -0.00295854 -0.01622007 -0.02424511  0.00170453 -0.01622007\n",
      " -0.01051013  0.01196934 -0.01622007  0.02183548 -0.01447533 -0.01622007\n",
      " -0.01721563 -0.01622007 -0.01622007 -0.01622007 -0.01622007 -0.01622007\n",
      " -0.01622007  0.01642573  0.02424511 -0.01622007 -0.01622007 -0.01622007\n",
      " -0.00295854 -0.01622007 -0.02162888  0.02424511]\n",
      "0.024245113492461084 -0.024245113492461057 0.04849022698492214\n",
      "0.6212255674623054 0.3787744325376947\n",
      "[0.41889964 0.41889964 0.47080012 0.41889964 0.45831883 0.41889964\n",
      " 0.47220972 0.43488251 0.43331727 0.45831883 0.37877443 0.41889964\n",
      " 0.47747214 0.41889964 0.41889964 0.41889964 0.49311914 0.41889964\n",
      " 0.37877443 0.48520731 0.41889964 0.37877443 0.50852266 0.41889964\n",
      " 0.44744935 0.55984669 0.41889964 0.60917741 0.42762333 0.41889964\n",
      " 0.41392187 0.41889964 0.41889964 0.41889964 0.41889964 0.41889964\n",
      " 0.41889964 0.58212863 0.62122557 0.41889964 0.41889964 0.41889964\n",
      " 0.48520731 0.41889964 0.39185558 0.62122557]\n",
      "[0.13197509 0.13197509 0.16302935 0.13197509 0.15511455 0.13197509\n",
      " 0.16394197 0.14102916 0.14012321 0.15511455 0.11106639 0.13197509\n",
      " 0.16738326 0.13197509 0.13197509 0.13197509 0.17794182 0.13197509\n",
      " 0.11106639 0.17254109 0.13197509 0.11106639 0.18883089 0.13197509\n",
      " 0.14845886 0.22896264 0.13197509 0.27380926 0.13686317 0.13197509\n",
      " 0.12924265 0.13197509 0.13197509 0.13197509 0.13197509 0.13197509\n",
      " 0.13197509 0.2484008  0.28580671 0.13197509 0.13197509 0.13197509\n",
      " 0.17254109 0.13197509 0.11760868 0.28580671]\n",
      "Sex                        \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Dizziness Type             \t Stroke               \t 0.97869 \t 0.16303 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Triggered by turning in bed? \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "SG Vertical N              \t Stroke               \t 0.97949 \t 0.15511 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "SG Horizintal N            \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "SG Torsional N             \t Stroke               \t 0.9786 \t 0.16394 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right Gaze Vertical N      \t Stroke               \t 0.98095 \t 0.14103 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right G Horizintal N       \t Stroke               \t 0.98104 \t 0.14012 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right G Torsional N        \t Stroke               \t 0.97949 \t 0.15511 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Vertical \t Stroke               \t 0.98422 \t 0.11107 \t [[0.   0.   0.   0.   0.99 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Horizint \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Torsiona \t Stroke               \t 0.97826 \t 0.16738 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Vertical N \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Torsional  \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Vertical \t Stroke               \t 0.97723 \t 0.17794 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Horizint \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Torsiona \t Stroke               \t 0.98422 \t 0.11107 \t [[0.   0.   0.   0.   0.99 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced V \t Stroke               \t 0.97776 \t 0.17254 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced T \t Stroke               \t 0.98422 \t 0.11107 \t [[0.   0.   0.   0.   0.99 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ve \t Stroke               \t 0.97619 \t 0.18883 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ho \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced To \t Stroke               \t 0.98017 \t 0.14846 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right Roll Induced Vertical  \t Stroke               \t 0.97257 \t 0.22896 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right Roll Induced Horizonta \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Right Roll Induced Torsional \t Stroke               \t 0.96885 \t 0.27381 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Roll Induced Vertical N \t Stroke               \t 0.98139 \t 0.13686 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Roll Induced Horizontal \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Left Roll Induced Torsional  \t Stroke               \t 0.98221 \t 0.12924 \t [[0.   0.   0.   0.   0.99 0.   0.  ]]\n",
      "Bow Induced Vertical Nystagm \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Bow Induced Torsional Nystag \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Lean Induced Vertical Nystag \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Lean Induced Torsional Nysta \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.97092 \t 0.2484 \t [[0.   0.   0.   0.02 0.98 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.9679 \t 0.28581 \t [[0.   0.   0.   0.01 0.98 0.   0.  ]]\n",
      "Head impulse test to the lef \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Gait                       \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Finger-nose-finger left hand \t Stroke               \t 0.97776 \t 0.17254 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Finger-nose-finger right han \t Stroke               \t 0.98191 \t 0.13198 \t [[0.   0.   0.   0.01 0.99 0.   0.  ]]\n",
      "Hearing loss right ear     \t Stroke               \t 0.98349 \t 0.11761 \t [[0.   0.   0.   0.   0.99 0.   0.  ]]\n",
      "Hearing loss left ear      \t Stroke               \t 0.9679 \t 0.28581 \t [[0.   0.   0.   0.01 0.98 0.   0.  ]]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[0.65481564 0.61602381 0.68344617 0.7476419  0.61602381 0.73210737\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.62026441 0.69659803 0.61602381 0.61602381 0.76187554\n",
      " 0.69659803 0.61602381 0.8134058  0.61602381 0.61602381 0.5666463\n",
      " 0.61602381 0.61602381 0.74905274 0.61602381 0.61602381 0.61602381\n",
      " 0.61602381 0.61602381 0.69782755 0.61602381 0.61602381 0.65481564\n",
      " 0.61602381 0.61602381 0.61602381 0.61602381 0.58023878 0.57905345\n",
      " 0.5666463  0.53089457 0.53089457 0.53089457]\n",
      "0.8134057988020574 0.5308945652761518\n",
      "0.6721501820391046\n",
      "[-0.01733455 -0.05612637  0.01129599  0.07549172 -0.05612637  0.05995719\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05188577  0.02444785 -0.05612637 -0.05612637  0.08972536\n",
      "  0.02444785 -0.05612637  0.14125562 -0.05612637 -0.05612637 -0.10550388\n",
      " -0.05612637 -0.05612637  0.07690256 -0.05612637 -0.05612637 -0.05612637\n",
      " -0.05612637 -0.05612637  0.02567737 -0.05612637 -0.05612637 -0.01733455\n",
      " -0.05612637 -0.05612637 -0.05612637 -0.05612637 -0.0919114  -0.09309674\n",
      " -0.10550388 -0.14125562 -0.14125562 -0.14125562]\n",
      "0.14125561676295284 -0.14125561676295284 0.2825112335259057\n",
      "1.0 0.0\n",
      "[0.43864121 0.30133047 0.5399842  0.7672167  0.30133047 0.7122294\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.31634085 0.5865376  0.30133047 0.30133047 0.81759925\n",
      " 0.5865376  0.30133047 1.         0.30133047 0.30133047 0.12654979\n",
      " 0.30133047 0.30133047 0.77221063 0.30133047 0.30133047 0.30133047\n",
      " 0.30133047 0.30133047 0.59088972 0.30133047 0.30133047 0.43864121\n",
      " 0.30133047 0.30133047 0.30133047 0.30133047 0.17466285 0.17046714\n",
      " 0.12654979 0.         0.         0.        ]\n",
      "[0.14322212 0.07717191 0.21269578 0.47147346 0.07717191 0.39187502\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.08314202 0.25240324 0.07717191 0.07717191 0.55685977\n",
      " 0.25240324 0.07717191 1.         0.07717191 0.07717191 0.02426225\n",
      " 0.07717191 0.07717191 0.47937327 0.07717191 0.07717191 0.07717191\n",
      " 0.07717191 0.07717191 0.25640625 0.07717191 0.07717191 0.14322212\n",
      " 0.07717191 0.07717191 0.07717191 0.07717191 0.03618377 0.03507441\n",
      " 0.02426225 0.         0.         0.        ]\n",
      "Sex                        \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Dizziness Type             \t Vestibular Neuritis L \t 0.477 \t 0.2127 \t [[0.1  0.05 0.01 0.03 0.19 0.6  0.02]]\n",
      "Triggered by turning in bed? \t Vestibular Neuritis L \t 0.3861 \t 0.47147 \t [[0.08 0.04 0.01 0.03 0.37 0.44 0.03]]\n",
      "SG Vertical N              \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "SG Horizintal N            \t Vestibular Neuritis L \t 0.4085 \t 0.39188 \t [[0.14 0.07 0.02 0.04 0.23 0.5  0.02]]\n",
      "SG Torsional N             \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Gaze Vertical N      \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right G Horizintal N       \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right G Torsional N        \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Up Gaze \"Induced\" Vertical N \t Vestibular Neuritis L \t 0.56174 \t 0.08314 \t [[0.04 0.02 0.01 0.02 0.18 0.72 0.02]]\n",
      "Up Gaze \"Induced\" Horizintal \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.04 0.02 0.01 0.01 0.36 0.54 0.02]]\n",
      "Up Gaze \"Induced\" Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Down Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Down Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.36538 \t 0.55686 \t [[0.13 0.07 0.02 0.02 0.29 0.42 0.05]]\n",
      "Down Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.45875 \t 0.2524 \t [[0.04 0.02 0.01 0.01 0.36 0.54 0.02]]\n",
      "Right Dix-Hallpike Induced V \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.28892 \t 1.0 \t [[0.06 0.04 0.01 0.02 0.56 0.27 0.05]]\n",
      "Right Dix-Hallpike Induced T \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Dix-Hallpike Induced Ve \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Dix-Hallpike Induced Ho \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.04 0.02 0.01 0.01 0.07 0.84 0.02]]\n",
      "Left Dix-Hallpike Induced To \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Roll Induced Vertical  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Right Roll Induced Horizonta \t Vestibular Neuritis L \t 0.38406 \t 0.47937 \t [[0.08 0.04 0.01 0.03 0.37 0.44 0.03]]\n",
      "Right Roll Induced Torsional \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Vertical N \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Horizontal \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Left Roll Induced Torsional  \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Bow Induced Vertical Nystagm \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Bow Induced Horizontal Nysta \t Vestibular Neuritis L \t 0.45703 \t 0.25641 \t [[0.04 0.02 0.01 0.02 0.36 0.54 0.02]]\n",
      "Bow Induced Torsional Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Lean Induced Vertical Nystag \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Lean Induced Horizontal Nyst \t Vestibular Neuritis L \t 0.51603 \t 0.14322 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Lean Induced Torsional Nysta \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Head impulse test to the lef \t Vestibular Neuritis L \t 0.56724 \t 0.07717 \t [[0.04 0.03 0.01 0.01 0.16 0.73 0.02]]\n",
      "Head impulse test to the rig \t Vestibular Neuritis L \t 0.61261 \t 0.03618 \t [[0.03 0.02 0.01 0.01 0.12 0.81 0.01]]\n",
      "Gait                       \t Vestibular Neuritis L \t 0.61408 \t 0.03507 \t [[0.02 0.01 0.   0.01 0.15 0.8  0.01]]\n",
      "Finger-nose-finger left hand \t Vestibular Neuritis L \t 0.62935 \t 0.02426 \t [[0.04 0.02 0.01 0.01 0.07 0.84 0.02]]\n",
      "Finger-nose-finger right han \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss right ear     \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Hearing loss left ear      \t Vestibular Neuritis L \t 0.67197 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "[[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n",
      "[[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "[['Vestibular Neuritis L']] [['Vestibular Neuritis L']] [5.]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.67232995 0.67232995 0.67232995 0.70069215 0.67232995 0.79985771\n",
      " 0.67232995 0.67232995 0.67232995 0.67232995 0.67232995 0.67232995\n",
      " 0.67232995 0.624559   0.75411547 0.67232995 0.67232995 0.76625817\n",
      " 0.624559   0.67232995 0.78238832 0.67232995 0.67232995 0.59699247\n",
      " 0.67232995 0.67232995 0.70168951 0.67232995 0.67232995 0.59699247\n",
      " 0.67232995 0.67232995 0.75494357 0.67232995 0.67232995 0.75548573\n",
      " 0.67232995 0.67232995 0.67232995 0.67232995 0.75138076 0.75548573\n",
      " 0.70359239 0.69891754 0.62541166 0.62947533]\n",
      "0.7998577078764277 0.5969924690013553\n",
      "0.6984250884388915\n",
      "[-0.02609514 -0.02609514 -0.02609514  0.00226706 -0.02609514  0.10143262\n",
      " -0.02609514 -0.02609514 -0.02609514 -0.02609514 -0.02609514 -0.02609514\n",
      " -0.02609514 -0.07386609  0.05569039 -0.02609514 -0.02609514  0.06783308\n",
      " -0.07386609 -0.02609514  0.08396323 -0.02609514 -0.02609514 -0.10143262\n",
      " -0.02609514 -0.02609514  0.00326442 -0.02609514 -0.02609514 -0.10143262\n",
      " -0.02609514 -0.02609514  0.05651849 -0.02609514 -0.02609514  0.05706064\n",
      " -0.02609514 -0.02609514 -0.02609514 -0.02609514  0.05295567  0.05706064\n",
      "  0.0051673   0.00049245 -0.07301343 -0.06894976]\n",
      "0.10143261943753623 -0.10143261943753623 0.20286523887507246\n",
      "1.0 0.0\n",
      "[0.37136714 0.37136714 0.37136714 0.5111752  0.37136714 1.\n",
      " 0.37136714 0.37136714 0.37136714 0.37136714 0.37136714 0.37136714\n",
      " 0.37136714 0.13588592 0.77451912 0.37136714 0.37136714 0.83437509\n",
      " 0.13588592 0.37136714 0.91388673 0.37136714 0.37136714 0.\n",
      " 0.37136714 0.37136714 0.51609158 0.37136714 0.37136714 0.\n",
      " 0.37136714 0.37136714 0.77860114 0.37136714 0.37136714 0.78127362\n",
      " 0.37136714 0.37136714 0.37136714 0.37136714 0.76103866 0.78127362\n",
      " 0.52547159 0.50242748 0.14008901 0.16012037]\n",
      "[0.1074739  0.1074739  0.1074739  0.19075726 0.1074739  1.\n",
      " 0.1074739  0.1074739  0.1074739  0.1074739  0.1074739  0.1074739\n",
      " 0.1074739  0.02644321 0.48306516 0.1074739  0.1074739  0.58827306\n",
      " 0.02644321 0.1074739  0.76064907 0.1074739  0.1074739  0.\n",
      " 0.1074739  0.1074739  0.19436846 0.1074739  0.1074739  0.\n",
      " 0.1074739  0.1074739  0.48965622 0.1074739  0.1074739  0.49401523\n",
      " 0.1074739  0.1074739  0.1074739  0.1074739  0.46186266 0.49401523\n",
      " 0.2014076  0.18446191 0.02744516 0.03239757]\n",
      "Sex                        \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Dizziness Type             \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Triggered by turning in bed? \t Vestibular Neuritis L \t 0.45302 \t 0.19076 \t [[0.04 0.02 0.01 0.01 0.36 0.54 0.02]]\n",
      "SG Vertical N              \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "SG Horizintal N            \t Stroke               \t 0.30923 \t 1.0 \t [[0.03 0.01 0.   0.01 0.83 0.1  0.01]]\n",
      "SG Torsional N             \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right Gaze Vertical N      \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right G Horizintal N       \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right G Torsional N        \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Up Gaze \"Induced\" Vertical N \t Vestibular Neuritis L \t 0.55615 \t 0.02644 \t [[0.02 0.01 0.   0.01 0.15 0.8  0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.3767 \t 0.48307 \t [[0.05 0.03 0.01 0.02 0.54 0.34 0.02]]\n",
      "Up Gaze \"Induced\" Torsional  \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Down Gaze \"Induced\" Horizint \t Vestibular Neuritis L \t 0.35897 \t 0.58827 \t [[0.13 0.07 0.02 0.02 0.29 0.42 0.05]]\n",
      "Down Gaze \"Induced\" Torsiona \t Vestibular Neuritis L \t 0.55615 \t 0.02644 \t [[0.02 0.01 0.   0.01 0.15 0.8  0.01]]\n",
      "Right Dix-Hallpike Induced V \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.33521 \t 0.76065 \t [[0.06 0.04 0.01 0.02 0.56 0.27 0.05]]\n",
      "Right Dix-Hallpike Induced T \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Dix-Hallpike Induced Ho \t Vestibular Neuritis L \t 0.5916 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Dix-Hallpike Induced To \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right Roll Induced Vertical  \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Right Roll Induced Horizonta \t Vestibular Neuritis L \t 0.45162 \t 0.19437 \t [[0.04 0.02 0.01 0.02 0.36 0.54 0.02]]\n",
      "Right Roll Induced Torsional \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Roll Induced Vertical N \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Left Roll Induced Horizontal \t Vestibular Neuritis L \t 0.5916 \t 0.0 \t [[0.01 0.01 0.01 0.   0.06 0.9  0.01]]\n",
      "Left Roll Induced Torsional  \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Bow Induced Vertical Nystagm \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.3755 \t 0.48966 \t [[0.04 0.03 0.01 0.02 0.54 0.34 0.02]]\n",
      "Bow Induced Torsional Nystag \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Lean Induced Vertical Nystag \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.37471 \t 0.49402 \t [[0.03 0.01 0.   0.01 0.66 0.29 0.01]]\n",
      "Lean Induced Torsional Nysta \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Alternate Cover (Moving from \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Head impulse test to the lef \t Vestibular Neuritis L \t 0.49228 \t 0.10747 \t [[0.03 0.02 0.01 0.01 0.29 0.64 0.01]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.38068 \t 0.46186 \t [[0.03 0.01 0.01 0.   0.59 0.33 0.02]]\n",
      "Gait                       \t Stroke               \t 0.37471 \t 0.49402 \t [[0.03 0.01 0.   0.01 0.66 0.29 0.01]]\n",
      "Finger-nose-finger left hand \t Vestibular Neuritis L \t 0.44896 \t 0.20141 \t [[0.04 0.02 0.01 0.01 0.39 0.52 0.01]]\n",
      "Finger-nose-finger right han \t Vestibular Neuritis L \t 0.4555 \t 0.18446 \t [[0.02 0.01 0.01 0.   0.4  0.53 0.02]]\n",
      "Hearing loss right ear     \t Vestibular Neuritis L \t 0.55503 \t 0.02745 \t [[0.02 0.01 0.01 0.   0.12 0.81 0.02]]\n",
      "Hearing loss left ear      \t Vestibular Neuritis L \t 0.54971 \t 0.0324 \t [[0.03 0.02 0.01 0.01 0.12 0.81 0.01]]\n",
      "[[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[0.1865194  0.1865194  0.23489092 0.20464056 0.1865194  0.1865194\n",
      " 0.27954039 0.1865194  0.19627449 0.20869779 0.1865194  0.1865194\n",
      " 0.19131231 0.1865194  0.1865194  0.1865194  0.23590572 0.1865194\n",
      " 0.22166088 0.1865194  0.1865194  0.25927966 0.1865194  0.1865194\n",
      " 0.19627449 0.1865194  0.1865194  0.22166088 0.1865194  0.1865194\n",
      " 0.24690492 0.1865194  0.1865194  0.1865194  0.1865194  0.1865194\n",
      " 0.1865194  0.21811921 0.1865194  0.1865194  0.1865194  0.1865194\n",
      " 0.22133457 0.1763444  0.1865194  0.26040104]\n",
      "0.27954038977261403 0.17634439780861785\n",
      "0.22794239379061593\n",
      "[-0.04142299 -0.04142299  0.00694853 -0.02330183 -0.04142299 -0.04142299\n",
      "  0.051598   -0.04142299 -0.0316679  -0.0192446  -0.04142299 -0.04142299\n",
      " -0.03663009 -0.04142299 -0.04142299 -0.04142299  0.00796332 -0.04142299\n",
      " -0.00628151 -0.04142299 -0.04142299  0.03133727 -0.04142299 -0.04142299\n",
      " -0.0316679  -0.04142299 -0.04142299 -0.00628151 -0.04142299 -0.04142299\n",
      "  0.01896253 -0.04142299 -0.04142299 -0.04142299 -0.04142299 -0.04142299\n",
      " -0.04142299 -0.00982318 -0.04142299 -0.04142299 -0.04142299 -0.04142299\n",
      " -0.00660783 -0.051598   -0.04142299  0.03245865]\n",
      "0.0515979959819981 -0.051597995981998074 0.10319599196399618\n",
      "0.7579899799099905 0.24201002009000966\n",
      "[0.29288504 0.29288504 0.53474265 0.38349085 0.29288504 0.29288504\n",
      " 0.75798998 0.29288504 0.34166048 0.40377699 0.29288504 0.29288504\n",
      " 0.31684956 0.29288504 0.29288504 0.29288504 0.53981661 0.29288504\n",
      " 0.46859243 0.29288504 0.29288504 0.65668633 0.29288504 0.29288504\n",
      " 0.34166048 0.29288504 0.29288504 0.46859243 0.29288504 0.29288504\n",
      " 0.59481265 0.29288504 0.29288504 0.29288504 0.29288504 0.29288504\n",
      " 0.29288504 0.4508841  0.29288504 0.29288504 0.29288504 0.29288504\n",
      " 0.46696087 0.24201002 0.29288504 0.66229324]\n",
      "[0.07392905 0.07392905 0.20856208 0.11339572 0.07392905 0.07392905\n",
      " 0.45718517 0.07392905 0.09384121 0.1237984  0.07392905 0.07392905\n",
      " 0.08334909 0.07392905 0.07392905 0.07392905 0.21256261 0.07392905\n",
      " 0.16160775 0.07392905 0.07392905 0.32373843 0.07392905 0.07392905\n",
      " 0.09384121 0.07392905 0.07392905 0.16160775 0.07392905 0.07392905\n",
      " 0.2600595  0.07392905 0.07392905 0.07392905 0.07392905 0.07392905\n",
      " 0.07392905 0.15053868 0.07392905 0.07392905 0.07392905 0.07392905\n",
      " 0.16056316 0.05603818 0.07392905 0.33011363]\n",
      "Sex                        \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Dizziness Type             \t Stroke               \t 0.9327 \t 0.20856 \t [[0.01 0.01 0.02 0.01 0.95 0.   0.  ]]\n",
      "Triggered by turning in bed? \t Stroke               \t 0.94878 \t 0.1134 \t [[0.   0.   0.01 0.01 0.97 0.   0.  ]]\n",
      "SG Vertical N              \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "SG Horizintal N            \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "SG Torsional N             \t Stroke               \t 0.90513 \t 0.45719 \t [[0.01 0.01 0.03 0.02 0.92 0.01 0.  ]]\n",
      "Right Gaze Vertical N      \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right G Horizintal N       \t Stroke               \t 0.95285 \t 0.09384 \t [[0.   0.   0.01 0.01 0.97 0.   0.  ]]\n",
      "Right G Torsional N        \t Stroke               \t 0.94675 \t 0.1238 \t [[0.   0.   0.01 0.01 0.97 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Vertical \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Horizint \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Torsiona \t Stroke               \t 0.95518 \t 0.08335 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Vertical N \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Torsional  \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Vertical \t Stroke               \t 0.93212 \t 0.21256 \t [[0.   0.   0.02 0.01 0.95 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Horizint \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Torsiona \t Stroke               \t 0.93999 \t 0.16161 \t [[0.   0.   0.01 0.01 0.96 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced V \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced T \t Stroke               \t 0.9182 \t 0.32374 \t [[0.   0.01 0.02 0.02 0.94 0.01 0.  ]]\n",
      "Left Dix-Hallpike Induced Ve \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ho \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced To \t Stroke               \t 0.95285 \t 0.09384 \t [[0.   0.   0.01 0.01 0.97 0.   0.  ]]\n",
      "Right Roll Induced Vertical  \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right Roll Induced Horizonta \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Right Roll Induced Torsional \t Stroke               \t 0.93999 \t 0.16161 \t [[0.   0.   0.01 0.01 0.96 0.   0.  ]]\n",
      "Left Roll Induced Vertical N \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Roll Induced Horizontal \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Left Roll Induced Torsional  \t Stroke               \t 0.92573 \t 0.26006 \t [[0.   0.01 0.02 0.02 0.95 0.   0.  ]]\n",
      "Bow Induced Vertical Nystagm \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Bow Induced Torsional Nystag \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Lean Induced Vertical Nystag \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Lean Induced Torsional Nysta \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.94188 \t 0.15054 \t [[0.   0.   0.01 0.01 0.96 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Head impulse test to the lef \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Gait                       \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Finger-nose-finger left hand \t Stroke               \t 0.94017 \t 0.16056 \t [[0.   0.   0.02 0.01 0.96 0.   0.  ]]\n",
      "Finger-nose-finger right han \t Stroke               \t 0.96188 \t 0.05604 \t [[0.   0.   0.01 0.   0.98 0.   0.  ]]\n",
      "Hearing loss right ear     \t Stroke               \t 0.95739 \t 0.07393 \t [[0.   0.   0.01 0.01 0.98 0.   0.  ]]\n",
      "Hearing loss left ear      \t Stroke               \t 0.9175 \t 0.33011 \t [[0.   0.01 0.02 0.02 0.94 0.01 0.  ]]\n",
      "[[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[0.16508731 0.16508731 0.50336451 0.22115062 0.16508731 0.16539967\n",
      " 0.16508731 0.16508731 0.16539967 0.16508731 0.16508731 0.23277613\n",
      " 0.16508731 0.1823336  0.19102006 0.16508731 0.16508731 0.48677033\n",
      " 0.37226494 0.16508731 0.25248673 0.16508731 0.16508731 0.17763844\n",
      " 0.16508731 0.16508731 0.30255146 0.16508731 0.16508731 0.1730046\n",
      " 0.16508731 0.16508731 0.16508731 0.16508731 0.16508731 0.16508731\n",
      " 0.16508731 0.17400039 0.26752852 0.19688934 0.16508731 0.16508731\n",
      " 0.15133043 0.15592788 0.18280557 0.25152882]\n",
      "0.5033645051836302 0.1513304275693398\n",
      "0.327347466376485\n",
      "[-0.16226016 -0.16226016  0.17601704 -0.10619685 -0.16226016 -0.1619478\n",
      " -0.16226016 -0.16226016 -0.1619478  -0.16226016 -0.16226016 -0.09457134\n",
      " -0.16226016 -0.14501386 -0.13632741 -0.16226016 -0.16226016  0.15942287\n",
      "  0.04491747 -0.16226016 -0.07486074 -0.16226016 -0.16226016 -0.14970903\n",
      " -0.16226016 -0.16226016 -0.02479601 -0.16226016 -0.16226016 -0.15434287\n",
      " -0.16226016 -0.16226016 -0.16226016 -0.16226016 -0.16226016 -0.16226016\n",
      " -0.16226016 -0.15334708 -0.05981894 -0.13045813 -0.16226016 -0.16226016\n",
      " -0.17601704 -0.17141959 -0.14454189 -0.07581865]\n",
      "0.1760170388071452 -0.1760170388071452 0.3520340776142904\n",
      "1.0 0.0\n",
      "[0.03907827 0.03907827 1.         0.19833361 0.03907827 0.03996557\n",
      " 0.03907827 0.03907827 0.03996557 0.03907827 0.03907827 0.23135743\n",
      " 0.03907827 0.08806867 0.11274371 0.03907827 0.03907827 0.95286202\n",
      " 0.6275941  0.03907827 0.28734803 0.03907827 0.03907827 0.07473143\n",
      " 0.03907827 0.03907827 0.42956361 0.03907827 0.03907827 0.06156838\n",
      " 0.03907827 0.03907827 0.03907827 0.03907827 0.03907827 0.03907827\n",
      " 0.03907827 0.06439707 0.33007627 0.12941619 0.03907827 0.03907827\n",
      " 0.         0.01305968 0.08940938 0.28462697]\n",
      "[0.00653663 0.00653663 1.         0.04271045 0.00653663 0.00669411\n",
      " 0.00653663 0.00653663 0.00669411 0.00653663 0.00653663 0.05262503\n",
      " 0.00653663 0.01588985 0.02114684 0.00653663 0.00653663 0.86137322\n",
      " 0.29232557 0.00653663 0.07184705 0.00653663 0.00653663 0.01320606\n",
      " 0.00653663 0.00653663 0.13796783 0.00653663 0.00653663 0.01066041\n",
      " 0.00653663 0.00653663 0.00653663 0.00653663 0.00653663 0.00653663\n",
      " 0.00653663 0.01119902 0.08884532 0.02492538 0.00653663 0.00653663\n",
      " 0.         0.00209993 0.01616561 0.07083648]\n",
      "Sex                        \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Dizziness Type             \t Stroke               \t 0.70336 \t 1.0 \t [[0.1  0.05 0.01 0.02 0.71 0.08 0.04]]\n",
      "Triggered by turning in bed? \t Stroke               \t 0.94027 \t 0.04271 \t [[0.01 0.   0.   0.   0.96 0.   0.03]]\n",
      "SG Vertical N              \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "SG Horizintal N            \t Stroke               \t 0.96644 \t 0.00669 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "SG Torsional N             \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right Gaze Vertical N      \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right G Horizintal N       \t Stroke               \t 0.96644 \t 0.00669 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right G Torsional N        \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Vertical \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Gaze \"Induced\" Horizint \t Stroke               \t 0.93389 \t 0.05263 \t [[0.01 0.01 0.   0.   0.95 0.01 0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Vertical N \t Stroke               \t 0.95926 \t 0.01589 \t [[0.   0.   0.   0.   0.98 0.01 0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.95532 \t 0.02115 \t [[0.01 0.   0.   0.   0.97 0.   0.  ]]\n",
      "Up Gaze \"Induced\" Torsional  \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Vertical \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Down Gaze \"Induced\" Horizint \t Stroke               \t 0.72165 \t 0.86137 \t [[0.13 0.03 0.   0.01 0.73 0.02 0.08]]\n",
      "Down Gaze \"Induced\" Torsiona \t Stroke               \t 0.83385 \t 0.29233 \t [[0.01 0.01 0.   0.   0.85 0.01 0.13]]\n",
      "Right Dix-Hallpike Induced V \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.92238 \t 0.07185 \t [[0.01 0.01 0.   0.01 0.94 0.01 0.03]]\n",
      "Right Dix-Hallpike Induced T \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ve \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Dix-Hallpike Induced Ho \t Stroke               \t 0.96132 \t 0.01321 \t [[0.   0.   0.   0.   0.98 0.   0.02]]\n",
      "Left Dix-Hallpike Induced To \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right Roll Induced Vertical  \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Right Roll Induced Horizonta \t Stroke               \t 0.88918 \t 0.13797 \t [[0.02 0.01 0.   0.   0.9  0.01 0.05]]\n",
      "Right Roll Induced Torsional \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Roll Induced Vertical N \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Left Roll Induced Horizontal \t Stroke               \t 0.9633 \t 0.01066 \t [[0.   0.   0.   0.   0.98 0.01 0.  ]]\n",
      "Left Roll Induced Torsional  \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Bow Induced Vertical Nystagm \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Bow Induced Torsional Nystag \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Lean Induced Vertical Nystag \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Lean Induced Torsional Nysta \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.96288 \t 0.0112 \t [[0.   0.   0.   0.   0.98 0.   0.01]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.91299 \t 0.08885 \t [[0.01 0.   0.   0.   0.93 0.01 0.04]]\n",
      "Head impulse test to the lef \t Stroke               \t 0.95256 \t 0.02493 \t [[0.01 0.   0.   0.   0.97 0.   0.01]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Gait                       \t Stroke               \t 0.96656 \t 0.00654 \t [[0.01 0.   0.   0.   0.98 0.   0.  ]]\n",
      "Finger-nose-finger left hand \t Stroke               \t 0.97188 \t 0.0 \t [[0.   0.   0.   0.   0.99 0.   0.01]]\n",
      "Finger-nose-finger right han \t Stroke               \t 0.97015 \t 0.0021 \t [[0.   0.   0.   0.   0.99 0.   0.01]]\n",
      "Hearing loss right ear     \t Stroke               \t 0.95905 \t 0.01617 \t [[0.   0.   0.   0.   0.98 0.   0.02]]\n",
      "Hearing loss left ear      \t Stroke               \t 0.92296 \t 0.07084 \t [[0.01 0.   0.   0.   0.94 0.02 0.02]]\n",
      "[[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "[['Stroke']] [['Stroke']] [4.]\n",
      "[0.44319531 0.44319531 0.44319531 0.40526342 0.44319531 0.49723468\n",
      " 0.44319531 0.44319531 0.44319531 0.44319531 0.44319531 0.44319531\n",
      " 0.44319531 0.43849903 0.44319531 0.44319531 0.44319531 0.43849903\n",
      " 0.45593029 0.44319531 0.45346035 0.44319531 0.44319531 0.5051806\n",
      " 0.44319531 0.44319531 0.40526342 0.44319531 0.44319531 0.53037386\n",
      " 0.44319531 0.44319531 0.51144842 0.44319531 0.44319531 0.49723468\n",
      " 0.44319531 0.44319531 0.44319531 0.44319531 0.44319531 0.44319531\n",
      " 0.44319531 0.44319531 0.40526342 0.50870243]\n",
      "0.5303738568522409 0.4052634173966034\n",
      "0.4678186371244222\n",
      "[-0.02462333 -0.02462333 -0.02462333 -0.06255522 -0.02462333  0.02941604\n",
      " -0.02462333 -0.02462333 -0.02462333 -0.02462333 -0.02462333 -0.02462333\n",
      " -0.02462333 -0.0293196  -0.02462333 -0.02462333 -0.02462333 -0.0293196\n",
      " -0.01188835 -0.02462333 -0.01435829 -0.02462333 -0.02462333  0.03736196\n",
      " -0.02462333 -0.02462333 -0.06255522 -0.02462333 -0.02462333  0.06255522\n",
      " -0.02462333 -0.02462333  0.04362978 -0.02462333 -0.02462333  0.02941604\n",
      " -0.02462333 -0.02462333 -0.02462333 -0.02462333 -0.02462333 -0.02462333\n",
      " -0.02462333 -0.02462333 -0.06255522  0.04088379]\n",
      "0.06255521972781874 -0.0625552197278188 0.12511043945563755\n",
      "0.8127760986390937 0.187223901360906\n",
      "[0.37688336 0.37688336 0.37688336 0.1872239  0.37688336 0.64708022\n",
      " 0.37688336 0.37688336 0.37688336 0.37688336 0.37688336 0.37688336\n",
      " 0.37688336 0.35340198 0.37688336 0.37688336 0.37688336 0.35340198\n",
      " 0.44055825 0.37688336 0.42820856 0.37688336 0.37688336 0.68680981\n",
      " 0.37688336 0.37688336 0.1872239  0.37688336 0.37688336 0.8127761\n",
      " 0.37688336 0.37688336 0.71814892 0.37688336 0.37688336 0.64708022\n",
      " 0.37688336 0.37688336 0.37688336 0.37688336 0.37688336 0.37688336\n",
      " 0.37688336 0.37688336 0.1872239  0.70441895]\n",
      "[0.11014164 0.11014164 0.11014164 0.03958952 0.11014164 0.31306187\n",
      " 0.11014164 0.11014164 0.11014164 0.11014164 0.11014164 0.11014164\n",
      " 0.11014164 0.09908498 0.11014164 0.11014164 0.11014164 0.09908498\n",
      " 0.14435013 0.11014164 0.13719568 0.11014164 0.11014164 0.35928243\n",
      " 0.11014164 0.11014164 0.03958952 0.11014164 0.11014164 0.54811663\n",
      " 0.11014164 0.11014164 0.3998279  0.11014164 0.11014164 0.31306187\n",
      " 0.11014164 0.11014164 0.11014164 0.11014164 0.11014164 0.11014164\n",
      " 0.11014164 0.11014164 0.03958952 0.38159518]\n",
      "Sex                        \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Stoke Risk (ABCD2 > 4)     \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Dizziness Type             \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Triggered by turning in bed? \t Stroke               \t 0.80413 \t 0.03959 \t [[0.01 0.01 0.   0.   0.92 0.05 0.01]]\n",
      "SG Vertical N              \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "SG Horizintal N            \t Stroke               \t 0.71017 \t 0.31306 \t [[0.03 0.01 0.01 0.   0.8  0.12 0.03]]\n",
      "SG Torsional N             \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right Gaze Vertical N      \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right G Horizintal N       \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right G Torsional N        \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Gaze \"Induced\" Vertical \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Gaze \"Induced\" Horizint \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Gaze \"Induced\" Torsiona \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Up Gaze \"Induced\" Vertical N \t Stroke               \t 0.77201 \t 0.09908 \t [[0.02 0.01 0.01 0.   0.88 0.07 0.01]]\n",
      "Up Gaze \"Induced\" Horizintal \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Up Gaze \"Induced\" Torsional  \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Down Gaze \"Induced\" Vertical \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Down Gaze \"Induced\" Horizint \t Stroke               \t 0.77201 \t 0.09908 \t [[0.02 0.01 0.01 0.   0.88 0.07 0.01]]\n",
      "Down Gaze \"Induced\" Torsiona \t Stroke               \t 0.75432 \t 0.14435 \t [[0.01 0.01 0.   0.   0.86 0.1  0.02]]\n",
      "Right Dix-Hallpike Induced V \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right Dix-Hallpike Induced H \t Stroke               \t 0.75686 \t 0.1372 \t [[0.04 0.01 0.   0.   0.86 0.07 0.01]]\n",
      "Right Dix-Hallpike Induced T \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Dix-Hallpike Induced Ve \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Dix-Hallpike Induced Ho \t Stroke               \t 0.70133 \t 0.35928 \t [[0.02 0.01 0.01 0.   0.8  0.06 0.1 ]]\n",
      "Left Dix-Hallpike Induced To \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right Roll Induced Vertical  \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Right Roll Induced Horizonta \t Stroke               \t 0.80413 \t 0.03959 \t [[0.01 0.01 0.   0.   0.92 0.05 0.01]]\n",
      "Right Roll Induced Torsional \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Roll Induced Vertical N \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Left Roll Induced Horizontal \t Stroke               \t 0.67258 \t 0.54812 \t [[0.03 0.01 0.01 0.   0.75 0.2  0.01]]\n",
      "Left Roll Induced Torsional  \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Bow Induced Vertical Nystagm \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Bow Induced Horizontal Nysta \t Stroke               \t 0.69428 \t 0.39983 \t [[0.02 0.01 0.01 0.   0.78 0.17 0.01]]\n",
      "Bow Induced Torsional Nystag \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Lean Induced Vertical Nystag \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Lean Induced Horizontal Nyst \t Stroke               \t 0.71017 \t 0.31306 \t [[0.03 0.01 0.01 0.   0.8  0.12 0.03]]\n",
      "Lean Induced Torsional Nysta \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Alternate Cover (Moving from \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Head impulse test to the lef \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Head impulse test to the rig \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Gait                       \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Finger-nose-finger left hand \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Finger-nose-finger right han \t Stroke               \t 0.7673 \t 0.11014 \t [[0.02 0.01 0.   0.   0.87 0.09 0.01]]\n",
      "Hearing loss right ear     \t Stroke               \t 0.80413 \t 0.03959 \t [[0.01 0.01 0.   0.   0.92 0.05 0.01]]\n",
      "Hearing loss left ear      \t Stroke               \t 0.69738 \t 0.3816 \t [[0.02 0.01 0.01 0.   0.78 0.17 0.01]]\n"
     ]
    }
   ],
   "source": [
    "for indx in range(len(test)):\n",
    "    for _ in range(5):\n",
    "        x = np.array([bm.sample_hidden(X_test[indx])])\n",
    "        baseline = clf.predict_proba(x)\n",
    "        print(np.round(baseline, 2))\n",
    "        pred_indx = clf.predict(x)\n",
    "        y_est = pred_indx.reshape((1, 1))\n",
    "        print(encoder.inverse_transform(y_test[indx].reshape(1, -1)), encoder.inverse_transform(y_est), pred_indx)\n",
    "\n",
    "    y_ests = []\n",
    "    d_prods = []\n",
    "    probs = []\n",
    "    for col in col_map:\n",
    "        mask = [False] * X_train.shape[1]\n",
    "        for i in col_map[col]['indx']:\n",
    "            mask[i] = True\n",
    "        hidden = np.array([bm.sample_hidden(X_test[indx], mask=mask, use_visible=False)])\n",
    "        masked_probs = clf.predict_proba(hidden)\n",
    "        pred_indx = clf.predict(hidden)\n",
    "        y_ests.append(pred_indx.reshape((1, 1)))\n",
    "        d_prods.append(np.dot(baseline, masked_probs.T)[0, 0])\n",
    "        probs.append(masked_probs)\n",
    "\n",
    "    d_prods2 = relative_feature_importance(np.array(d_prods))\n",
    "\n",
    "    for col, y_est, d_prod, d2, masked_probs in zip(col_map, y_ests, d_prods, d_prods2, probs):\n",
    "        print('{:26s} \\t {:20s} \\t'.format(col[:28], encoder.inverse_transform(y_est)[0, 0]), np.round(d_prod, 5), '\\t', np.round(d2, 5), '\\t', np.round(masked_probs, 2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 100,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.86907959, 0.40101544, 0.67149804, 0.28962556, 0.71565635,\n",
       "       0.15401548, 0.88231164, 0.        , 0.49329102, 0.24337936,\n",
       "       0.0964698 , 0.0729    , 0.42744271, 0.06126908, 0.56471916,\n",
       "       0.1539    , 0.60999267, 0.        , 0.11981   , 0.54306965,\n",
       "       0.11333675, 0.44256977, 0.        , 0.22955108, 0.77866036,\n",
       "       0.47663013, 0.18756615, 0.73543828, 0.6208789 , 0.54718713,\n",
       "       0.34012398, 0.06561   , 0.        , 0.0819    , 0.64829683,\n",
       "       0.36811224, 0.507942  , 0.06561   , 0.68071033, 0.14430952,\n",
       "       0.46380587, 0.0081    , 0.        , 0.42482634, 0.23902617,\n",
       "       0.08019   , 0.51083313, 0.53137805, 0.19869855, 0.15121291,\n",
       "       0.        , 0.41614105, 0.73575113, 0.75231031, 0.        ,\n",
       "       0.43962174, 0.07656812, 0.67507108, 0.17536676, 0.59064732,\n",
       "       0.        , 0.42444025, 0.        , 0.1004355 , 0.00280602,\n",
       "       0.11286471, 0.02500354, 0.25326061, 0.6511529 , 0.65032054,\n",
       "       0.05839422, 0.48535887, 0.        , 0.        , 0.7066939 ,\n",
       "       0.26640411])"
      ]
     },
     "execution_count": 100,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "bm.visible_bias"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}