examples/LIME_Education_ex_notebook.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Predicting mid-career salary based on education data\n",
    "\n",
    "For this example we employ three datasets:\n",
    "\n",
    "- One containing data sourced form the Wall Street Journal that analyzes salaries by college, region and academic major.\n",
    "- A dataset compiled and shared through Kaggle, containing attributes related to many of the same colleges and universities as in the dataset above.\n",
    "- Data from the College Scorecard published by the U.S. Department of Education.\n",
    "\n",
    "Using this dataset we fit a multiclass logistic regression model that classifies students into likely salary brackets they will reach mid-career, depending on which school, major and other factors are applicable to each."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sys\n",
    "sys.path.insert(0, '../LIMEaid/controller')\n",
    "sys.path.insert(0, '../LIMEaid/model')\n",
    "sys.path.insert(0, '../LIMEaid/view')\n",
    "import fit_sklearn_models as fsm\n",
    "import LIMEaid as la\n",
    "import LIMEdisplay as ld\n",
    "import load_college_dataset as gcd\n",
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "from sklearn import preprocessing\n",
    "from sklearn import tree\n",
    "\n",
    "# Number of perturbed samples to be generated.\n",
    "n = 100000\n",
    "# Number of bins for the histograms of continous attributes.\n",
    "num_bins = 25"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The functions below merge and prepare the data, after uploading it, and leave it ready for us to fit the logistic regression model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<_io.TextIOWrapper name='C:\\\\Users\\\\Castor18\\\\OneDrive\\\\MSDS_UW\\\\Data515_Spr19_Sw_Eng\\\\myproject\\\\LIMEaid\\\\LIMEaid\\\\model\\\\../data/Most-_Recent-_Cohorts-_Scorecard-_Elements.csv' mode='r' encoding='cp1252'>\n"
     ]
    }
   ],
   "source": [
    "college = gcd.load_college_dataset()\n",
    "features = list(college)[1:]\n",
    "salary_class = np.array(college.iloc[:, 0])\n",
    "college = np.array(college.iloc[:, 1:])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next we fit the logistic regression model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<_io.TextIOWrapper name='C:\\\\Users\\\\Castor18\\\\OneDrive\\\\MSDS_UW\\\\Data515_Spr19_Sw_Eng\\\\myproject\\\\LIMEaid\\\\LIMEaid\\\\model\\\\../data/Most-_Recent-_Cohorts-_Scorecard-_Elements.csv' mode='r' encoding='cp1252'>\n"
     ]
    }
   ],
   "source": [
    "clf = fsm.fit_multiclass_logistic_regression()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Interpreting the output of the Logistic Regression model using LIME.\n",
    "\n",
    "We now take the dataset, attributes only, and form them we generate n random samples with similar distributions in the values of each attribute."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Note that, unlike the Iris example where all attributes had continuous\n",
    "# values, in this dataset the first attribute has discrete values. So, we\n",
    "# call lime_sample setting the \"continuous\" parameter, second one in the\n",
    "# call, to False.\n",
    "perturbed_samples = la.lime_sample(n, False, college[:, 0], num_bins)\n",
    "# The for loop calls lime_sample for each of the remaining attributes,\n",
    "# setting the \"continuous\" parameter to True.\n",
    "for j in range(1, college.shape[1]):\n",
    "    array = college[:, j]\n",
    "    output = la.lime_sample(n, True, array, num_bins)\n",
    "    perturbed_samples = np.vstack((perturbed_samples, output))\n",
    "perturbed_samples = np.transpose(perturbed_samples)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Once random samples with the right distributions for each attribute have been generated, we provide them as input to the Logistic Regression model we fitted earlier, and obtain the classification for each."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "class_perturb_samples = clf.predict(perturbed_samples)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We select a single instance from the dataset. This is the instance we will try to interpret using LIME."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Normalizing the data.\n",
    "college_norm = preprocessing.scale(college)\n",
    "# Selecting the instance to interpret.\n",
    "inst_num = np.round(np.random.uniform(0, college_norm.shape[0], 1))\n",
    "inst_num = inst_num[0].astype(int)\n",
    "# x is the selected instance, and x_class is the class assigned\n",
    "# by the decision tree. \n",
    "x = college_norm[inst_num, :]\n",
    "x_class = salary_class[inst_num]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Calling LIME function to get intepretation."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We now fit the LIME linear model to get the coefficients and\n",
    "# intercept, as well as the weight of each random sample, \n",
    "# based on its L2 distance to the instance that is being\n",
    "# interpreted.\n",
    "lime_beta, lime_int, lime_weigh = la.lime_fit(x,\n",
    "                                              x_class,\n",
    "                                              perturbed_samples,\n",
    "                                              class_perturb_samples)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Intepreting the results\n",
    "\n",
    "Below we present the results obtained from the LIME linear regression model, and identify those attributes that played a significant role in the classification that was assigned to each instance by the Logistic Regression classifier. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Instance to be interpreted:\n",
      "Feature:  HCM2 \tvalue:  0.0 \tnormalized value:  0.0\n",
      "Feature:  SATVR25 \tvalue:  610.0 \tnormalized value:  0.8279109045530632\n",
      "Feature:  SATVR75 \tvalue:  710.0 \tnormalized value:  0.827887130425389\n",
      "Feature:  SATMT25 \tvalue:  620.0 \tnormalized value:  0.7527182896284788\n",
      "Feature:  SATMT75 \tvalue:  720.0 \tnormalized value:  0.7617365530839537\n",
      "Feature:  SATWR25 \tvalue:  630.0 \tnormalized value:  1.0394908779151995\n",
      "Feature:  SATWR75 \tvalue:  710.0 \tnormalized value:  0.7710461150749859\n",
      "Feature:  ACTCM25 \tvalue:  28.0 \tnormalized value:  0.9532444604150984\n",
      "Feature:  ACTCM75 \tvalue:  32.0 \tnormalized value:  0.9689035793685704\n",
      "Feature:  ACTEN25 \tvalue:  28.0 \tnormalized value:  1.1968140340296447\n",
      "Feature:  ACTEN75 \tvalue:  33.0 \tnormalized value:  1.1624138601596006\n",
      "Feature:  ACTMT25 \tvalue:  27.0 \tnormalized value:  1.1435900194808717\n",
      "Feature:  ACTMT75 \tvalue:  32.0 \tnormalized value:  1.1577354446778019\n",
      "Feature:  ACTWR25 \tvalue:  7.408163265306122 \tnormalized value:  -1.934463152455155e-15\n",
      "Feature:  ACTWR75 \tvalue:  8.959183673469388 \tnormalized value:  0.0\n",
      "Feature:  PCIP01 \tvalue:  0.0 \tnormalized value:  -0.3061614792206546\n",
      "Classification:  1\n",
      "\n",
      "Significant coefficients from LIME adjusted linear model:\n",
      "Feature:  ACTCM25 \tCoefficient:  0.7250634695611433\n",
      "Feature:  PCIP01 \tCoefficient:  0.01636399071349308\n",
      "Intercept:  0.5031131375596364\n"
     ]
    }
   ],
   "source": [
    "# Print output of LIME results.\n",
    "# First list the attributes and values for the instance\n",
    "# we are looking at.\n",
    "print(\"Instance to be interpreted:\")\n",
    "for j in range(0, len(lime_beta)):\n",
    "    print(\"Feature: \", features[j], \"\\tvalue: \",\n",
    "          college[inst_num, j], \"\\tnormalized value: \",\n",
    "          college_norm[inst_num, j])\n",
    "print(\"Classification: \", x_class)\n",
    "# Second we list the attributes that LIME has\n",
    "# identified as significant in this case, along with\n",
    "# the corresponding LIME coefficients and the LIME\n",
    "# intercept.\n",
    "print(\"\\nSignificant coefficients from LIME adjusted\"\n",
    "      \" linear model:\")\n",
    "significant_attributes = 0\n",
    "for j in range(0, len(lime_beta)):\n",
    "    if(lime_beta[j] != 0):\n",
    "        significant_attributes = np.append(significant_attributes, j)\n",
    "        print(\"Feature: \", features[j],\n",
    "              \"\\tCoefficient: \", lime_beta[j])\n",
    "significant_attributes = significant_attributes[1:]\n",
    "print(\"Intercept: \", lime_int)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "If the number of significant attributes in the interpretation of a particular instance is equal to two, we plot the random samples and the instance that is being interpreted on a cathesian plane. In some cases, classes can be separated for the most part by looking at the two selected attributes. Note that those random samples that were classified differently from the instance we are interpreting are merged into a single out-of-class representation.\n",
    "\n",
    "Finally, we plot the output of the fitted LIME Linear Regression model for each sample, against its class, as yielded by the Logistic Regression model and see that, in most cases, the classes can be differentiated using this model only."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAVYAAAJcCAYAAACi3Ix4AAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4yLCBodHRwOi8vbWF0cGxvdGxpYi5vcmcvOIA7rQAAIABJREFUeJzt3XmYZHV97/H3RwYQEUUdRG2hR0dcCHEdXK6J4pbgirnECGKMXpVE9LrHoF73uBuNS0THgOKGEscoGoz7uGtklKBoVERKWRRR2UQdGb/3jzo9Fm13T/XMr7umut6v56mnz36+Vaf707/6nVOnUlVIktq52qgLkKSVxmCVpMYMVklqzGCVpMYMVklqzGCVpMYMVi2bJM9K8q+jrmNQko1JHjPqOpbDJD3XUVs16gI0vCTnAPsCW4DLgf8EnlBVl4+yrmFV1UtGXYO0HGyxjp8HVtU1gdsCtwOeuRQ7SbLLUmxX7SSxYbSTMljHVFX9GPgo/YAFIMnuSV6V5IdJfpLkTUn2GJj/jCQXJDk/yWOSVJKbdfPeluS4JKcm+SVwj4W2l2R1kg8nuTjJz5N8LsnVunn/kOS8JJcl+U6Se3XTn5/knQP1PCjJmd02Nia51cC8c5I8PckZSS5J8t4kV5/9OnQ1XpzkoIFp+yT5VZLrJ7lOV+dPk/yiG77xXK/pHPWt6V6jVd34tZMc372G5yX5x5l/QEluluQzXa0XJXnvPPuY2ebR3XG4IMnTBuZfLcmxSb6f5GdJTk5y3VnrPjrJD4FPzbOPw5KcnuTSbjuHzrHM2iSf6vZxUZJ3Jdl7YP58x/COSU7rtv2TJK+eq4ZJZ7COqS4c7gucNTD55cDN6YftzYAp4Lnd8ocCTwXu3c27+xybfRjwYmAv4PMLbQ94GnAusA/97olnAZXkFsATgIOrai/gz4Fz5qj/5sBJwJO7bZwKfCjJbgOL/RVwKHAT4NbAI2dvp6p+A7wfOHLWep+pqgvp/46/FZgG9gd+Bbxhjuc+jBOBK+m/FrcD/gyY6bN8EfAx4DrAjYHXb2Nb9wAO6LZxbJJ7d9OfCDyY/vG5EfAL4F9mrXt34Fb0X9urSHJH4O3A3wN7A3djjtcfCPDSbh+3AvYDnt9tY6Fj+FrgtVV1LWAtcPI2nudkqiofY/Kg/8t9OXAZUMAngb27eQF+CawdWP4uwA+64ROAlw7Mu1m3jZt1428D3j4wf1vbeyHwwZn1Z233QvoBvuusec8H3tkNPwc4eWDe1YDzgEMGnuvDB+a/AnjTPK/LvYGzB8a/ADxinmVvC/xiYHwj8JjZ9XXja7rXaBX9fx6/AfYYmH8k8Olu+O3AeuDG2ziGM9u85azndnw3/G3gXgPzbgj8tqthZt2bLrD9NwOvmWfe1uc6x7wHA18f4hh+FngBsHrUfw8788MW6/h5cPVbEYcAtwRWd9P3Aa4BbOreGl9M/+TWPt38GwE/GtjO4PBc07a1vVfSby1/LMnZSY4FqKqz6LdCnw9cmOQ9SW40x75uBPRmRqrqd93+pwaW+fHA8BXANefYDvTfEu+R5E5JpumH578DJLlGkjcn6SW5lH4w7J3F9yFPA7sCFwy8Hm8Grt/Nfwb9f0b/1XVv/J9tbG/wte7Rfz1m9vPvA/v4Nv2TlfvOs+5s+wHf39aT6bpJ3tO93b8UeCfd79I2juGj6b+L+Z8kX03ygG3taxIZrGOqqj5Dv5X5qm7SRfTf5v5RVe3dPa5d/RNdABfQf4s6Y7+5NjswvOD2quqyqnpaVd0UeCDw1Jl+uKp6d1X9Cf2QKPpdCrOd380HIEm6ms4b/lXoiu6H8sn0W5APAz5cVZd1s58G3AK4U/Xfvt5tZpdzbOqX9P+ZzLjBwPCP6LdYVw+8Hteqqj/qavhxVT22qm4E/C3wxnT91/MYfP33p/96zOznvgP72Luqrl5Vg6/LQrek+xH9t+jb8tJuO7fuXpeHM/CazHcMq+p7VXUk/X8oLwfel2TPIfY3UQzW8fbPwH2S3LYLl7cAr0lyfYAkU0lm+uFOBh6V5FZJrsHv+0rntK3tJXlAd8ImwKX0W1VbktwiyT2T7A78mn44b5ljFycD909yryS70g/A3wBf3M7X4t3AQ4GjuuEZe3U1XNydBHreAts4Hbhbkv2TXJuBKy6q6gL6faj/lORa3UmmtUnuDpDkIQMnxX5BP4zmet4zntO1pv8IeBQwc7LrTcCLu5b3zIm4w4Z5ATrH0z/O9+pqnEpyyzmW24t+t9LFSabo98nS7XPeY5jk4Un26X4/Lu5WWeh5TiSDdYxV1U/p9+09p5v0D/Tfnn+5e3v3CfqtNarqI8DrgE93y3ypW+c3C+xi3u3RP/HyCfp/nF8C3lhVG4HdgZfRb/H+mH7L5llz1P4d+q2k13fLPpD+pWSbF/MaDGzvK/RbnDcCPjIw65+BPbp9fJl+d8Z82/g4/YA7A9gEfHjWIo8AdgO+RT8830e/DxTgYOArSS4HTgGeVFU/WKDkz9B/bT8JvKqqPtZNf223/seSXNbVfKcFtjP7OfwX/aB+DXBJt5/pORZ9AXD7bpn/oH8CcMZCx/BQ4Mzueb4WOKKqfg2Q5PIkfzpsrStZug5pTZj0L236JrB7VV056nomRZI1wA/onxTydV+hbLFOkCR/kWS3JNeh3z/2If+4pfYM1snyt8BP6Z813gI8brTlSCuTXQGS1JgtVklqbMXdxGH16tW1Zs2aRa+3efNmdtttt20vqBXF4z55tveYb9q06aKq2mfbS67AYF2zZg2nnXbaotfr9XpMT891VYpWMo/75NneY56kt+2l+uwKkKTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJamzFff21JM1lYzYCsHrDajau6Q8fUocsyb5ssUpa8WZCddjpO8pglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGRhasSfZL8ukk305yZpInzbFMkrwuyVlJzkhy+1HUKkmLsWqE+74SeFpVfS3JXsCmJB+vqm8NLHNf4IDucSfguO6nJO20RtZiraoLqupr3fBlwLeBqVmLHQa8vfq+DOyd5IbLXKokLcooW6xbJVkD3A74yqxZU8CPBsbP7aZdMGv9o4GjAaampuj1eouuYfPmzdu1nsabx30yrN6weuvwqrWrrjK+FMd/5MGa5JrABuDJVXXp7NlzrFJ/MKFqPbAeYN26dTU9Pb3oOnq9Htuznsabx30ybFyzcevw6g2ruejwi7aOH1QHNd/fSK8KSLIr/VB9V1W9f45FzgX2Gxi/MXD+ctQmSdtrlFcFBDge+HZVvXqexU4BHtFdHXBn4JKqumCeZSVppzDKroC7An8NfCPJ6d20ZwH7A1TVm4BTgfsBZwFXAI8aQZ2StCgjC9aq+jxz96EOLlPA45enIklqw09eSVJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqukFe9W77zVoqbvqJHfNlCSltq+R+0LwNnPPhuA3ad356YvvunW6a0ZrJImwr5H7cu+R+1Lr9fjoHPa34N1kF0BktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSY9wqQNBE2ZiMAqzesZuOa/vAhdciS7MsWq6QVbyZUh52+owxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWrMYJWkxgxWSWpspMGa5IQkFyb55jzzD0lySZLTu8dzl7tGSVqsVSPe/9uANwBvX2CZz1XVA5anHEnacSNtsVbVZ4Gfj7IGSSvfIXXIoqbvqFG3WIdxlyT/DZwPPL2qzpy9QJKjgaMBpqam6PV6i97J5s2bt2s9jTeP++S4yTk3AfrHfGZ4qY79zh6sXwOmq+ryJPcDPgAcMHuhqloPrAdYt25dTU9PL3pHvV6P7VlP483jPnmW45jv1FcFVNWlVXV5N3wqsGuS1SMuS5IWtFMHa5IbJEk3fEf69f5stFVJ0sJG2hWQ5CTgEGB1knOB5wG7AlTVm4C/BB6X5ErgV8ARVVUjKleShjLSYK2qI7cx/w30L8eSpLGxU3cFSNI4MlglqTGDVZIaM1glqTGDVZIaM1glqbGd/SOtktTExmwEYPWG1Wxc0x9eqpuw2GKVtOLNhOqw03eUwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsmkhJ/7Fp0++HpVYMVk2c+ULUcFUrBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqskNWawSlJjBqsm0pG8ix+whjuwiR+whiN516hL0gqyatQFSMvtSN7FWziaPbmCHrCGHm/h6G7uUaMsTSuELVZNnJfwbPbkiqtM25MreAnPHlFFWmkMVk2c/fnhoqZLi2WwauL8kP0XNV1aLINVE+dZvJhfco2rTPsl1+BZvHhEFWmlMVg1cU7iKB7Les5hGoBzmOaxrOckT1ypEa8K0EQ6iaM4iaPYQI/DOWfU5WiFscUqSY0ZrJo4VYubrvF3SB2yqOk7yq4ATaSZEO31DNRJMROivV6Pg+qgJd2XLVZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxglaTGDFZJasxg1URK+o9Nm34/LLVisGrizBeihqtaMVglqTGDVZIaM1glqbGJD9ZjjjmGVatWsWnTJlatWsUxxxwz6pIkjbmJ/gaBY445huOOO27r+JYtW7aOv/GNbxxVWZLG3ES3WNevX7+o6ZI0jJEGa5ITklyY5JvzzE+S1yU5K8kZSW7fcv9btmxZ1HRJGsaoW6xvAw5dYP59gQO6x9HAcQssu2iZ58LF+aZL0jBGGqxV9Vng5wsschjw9ur7MrB3khs23P+ipkvSMHb2k1dTwI8Gxs/tpl0wuFCSo+m3aJmamqLX6w218Q0bNmwdXrt27VXGh92Gxs/AYWbt2s1s2PD7Y+1hX/k2b9685H/fO3uwzvWe/A+ak1W1HlgPsG7dupqenh5q42vWrNk6vGHDBg4//PDBbS6uUo2NgcPOhg09Dj/8978vHvaVr9frMWxGbK9R97Fuy7nAfgPjNwbOH1EtkjSUnT1YTwEe0V0dcGfgkqq6YFsrSdIojbQrIMlJwCHA6iTnAs8DdgWoqjcBpwL3A84CrgAeNZpKJWl4Iw3WqjpyG/MLePwylSNJTezsXQGSNHYMVklqzGCVpMYMVklqzGCVpMa2eVVAkrsCzwemu+VD/4T9TZe2NEkaT8NcbnU88BRgE+D99CRpG4YJ1kuq6iNLXokkrRDDBOunk7wSeD/wm5mJVfW1JatKksbYMMF6p+7nuoFpBdyzfTmSNP62GaxVdY/lKESSVoptXm6V5NpJXp3ktO7xT0muvRzFSdI4GuY61hOAy4C/6h6XAm9dyqIkaZwN08e6tqoOHxh/QZLTl6ogSRp3w7RYf5XkT2ZGug8M/GrpSpKk8TZMi/VxwIldv2rof6vqI5eyKEkaZ8NcFXA6cJsk1+rGL13yqiRpjM0brEkeXlXvTPLUWdMBqKpXL3FtkjSWFmqx7tn93Gs5CpGklWLeYK2qN3c/X7B85UjS+BvmAwKvSHKtJLsm+WSSi5I8fDmKk6RxNMzlVn/WnbB6AHAucHPg75e0KkkaY8ME667dz/sBJ1XVz5ewHkkae8Ncx/qhJP9D/0MBxyTZB/j10pYlSeNrmy3WqjoWuAuwrqp+C/wSOGypC5OkcTXMyauHAFdW1ZYk/w94J3CjJa9MksbUMH2sz6mqy7r7Bfw5cCJw3NKWJUnja5hgnfkCwfsDx1XVB4Hdlq6k5VNVi5qulWG+w+thVyvDnLw6L8mbgXsDL0+yO8MF8liYCdFer2egTpCZQ93rGahqb5iA/Cvgo8ChVXUxcF28jlWS5jXMVQFXABcCM/dkvRL43lIWJUnjbJirAp4H/APwzG7SrvSvDJAkzWGYroC/AB5E//pVqup8vOOVJM1rmGDdXP2zOgWQZM9tLC9JE22YYD25uypg7ySPBT4BvGVpy5Kk8TXMV7O8Ksl96H/t9S2A51bVx5e8MkkaUwsGa5JdgI9W1b0Bw1SShrBgV0BVbQGu6L6hVZI0hGE+efVr4BtJPk53ZQBAVT1xyaqSpDE2zMmr/wCeA3wW2DTwWBGmpqZIwqZNm0jC1NTUqEuSNOaGOXl14nIUMgpTU1Ocf/75V5l2/vnnMzU1xXnnnTeiqiSNu20Ga5Jv0F3DOuAS4DTgH6vqZ0tR2HKYHarbmi5Jwximj/Uj9G8d+O5u/Agg9MP1bcADl6QySRpTwwTrXavqrgPj30jyhaq6q1+DLUl/aJiTV9dMcqeZkSR3BK7ZjV65JFVJ0hgbpsX6GOCEJDNhehnwmO6eAS9dssokaUwNc1XAV4E/7j4kkO5m1zNOXrLKJGlMDXM/1n2THA+8p6ouTnJgkkcvQ22SNJaG6WN9G/2vZpn5yuvvAk9eqoIkadwNE6yrq+pk4HcAVXUlv//mVknSLMME6y+TXI/f3+j6zvSvYZUkzWGYqwKeCpwCrE3yBWAf4CFLWpUkjbFhrgr4WpK707/JdYDvVNVvl7wySRpTw3QFUFVXVtWZVfVN4JDuFoKSpDnMG6xJ7pnku0kuT/LO7jKr04CXAcctX4mSNF4WarH+E3A0cD3gfcCXgXdU1R2q6v3LUZwkjaOF+lirqjZ2wx9I8tOqeu0y1CRJY22hYN07yf8eGM/guK1WSZrbQsH6Ga56r9XB8QIMVkmaw7zBWlWPWs5CJGmlGOpyK0nS8AxWSWrMYJWkxhb6gMAzBoYfMmveS5ayKEkaZwu1WI8YGH7mrHmHLkEtkrQiLBSsmWd4rnFJUmehYK15hucalyR1FvqAwG2SXEq/dbpHN0w3fvUlr0ySxtRCHxDYZTkLkaSVYt5gTXLdhVasqp+3L0eSxt9CXQEXAecCV3bjgyesCrjpUhUlSeNsoWB9PXAI8AXgJODzVeVJK0nahnmvCqiqJwG3Bf4N+Gvg60lekeQmrXae5NAk30lyVpJj55j/yCQ/TXJ693hMq31L0lJZ8MsEuxbqp5N8nf4HBl4EfA94y47uOMkuwL8A96Hf5fDVJKdU1bdmLfreqnrCju5PkpbLQiev9gQOAx5K/yuv3w/cvqp+1GjfdwTOqqqzu/29p9vf7GCVpLGyUIv1Qvqt05OAs+ifsDo4ycHQ5BsEpoDBkD4XuNMcyx2e5G7Ad4GnzBXsSY6m//1cTE1N0ev1hipgw4YNW4fXrl17lfFht6HxtnnzZo/1hFmOY75QsP4b/TC9ZfcY1OIbBOb6WOzsk2MfAk6qqt8k+TvgROCef7BS1XpgPcC6detqenp6qALWrFmzdXjDhg0cfvjhg9scahsab71ej2F/X7QyLMcxX+gDAo9c0j33W6j7DYzfGDh/Vg0/Gxh9C/DyJa5JknbYQn2sT11oxap69Q7u+6vAAd1VBufRPzn2sFk13LCqLuhGHwR8ewf3KUlLbqGugL0WmLfD75Or6sokTwA+CuwCnFBVZyZ5IXBaVZ0CPDHJg+h/SOHnwCN3dL+StNQW6gp4wXzzkjy5xc6r6lTg1FnTnjsw/Ez+8F6wkrRT296vZlmwm0CSJtn2Bqs3upakeWxvsHotkiTNY6GrAi5j7gANsMeSVSRJY26hk1cLXRUgSZrH9nYFSJLmYbBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMGqyQ1ZrBKUmMTHaxVtajpkjSMVaMuYNRmQrTX6xmokpqY6BarJC0Fg1WSGjNYJakxg1WSGjNYJakxg1WSGjNYJakxg1WSGjNYJakxg1WSGpv4YE1CEjZt2rR1WJJ2xEQH63wharhK2hETHayStBQMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMYMVklqzGCVpMZGGqxJDk3ynSRnJTl2jvm7J3lvN/8rSdYsf5WStDgjC9YkuwD/AtwXOBA4MsmBsxZ7NPCLqroZ8Brg5ctbpSQt3ihbrHcEzqqqs6tqM/Ae4LBZyxwGnNgNvw+4V/xCKkk7uVUj3PcU8KOB8XOBO823TFVdmeQS4HrARYMLJTkaOBpgamqKXq83VAEbNmzYOrx27dqrjA+7DY23zZs3e6wnzHIc81EG61wtz9qOZaiq9cB6gHXr1tX09PRQBaxZs2br8IYNGzj88MMHtznUNjTeer0ew/6+aGVYjmM+yq6Ac4H9BsZvDJw/3zJJVgHXBn6+LNVJ0nYaZbB+FTggyU2S7AYcAZwya5lTgL/phv8S+FTZlJS0kxtZV0DXZ/oE4KPALsAJVXVmkhcCp1XVKcDxwDuSnEW/pXrEqOqVpGGNso+VqjoVOHXWtOcODP8aeMhy1yVJO8JPXklSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYxMdrPN9fZZfqyVpR4z0q1l2BjMh2uv1DFRJTUx0i1WSloLBKkmNGayS1JjBKkmNGayS1JjBKkmNGayS1JjBKkmNGayS1JjBKkmNTXywJiEJmzZt2josSTtiooN1vhA1XCXtiIkOVklaCgarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYyMJ1iTXTfLxJN/rfl5nnuW2JDm9e5yy3HVK0vYYVYv1WOCTVXUA8MlufC6/qqrbdo8HtS7iwAMPXNR0SRrGqIL1MODEbvhE4MGjKOLMM8/8gxA98MADOfPMM0dRjqQVIlW1/DtNLq6qvQfGf1FVf9AdkORK4HTgSuBlVfWBebZ3NHA0wNTU1B2+8IUvLLqmzZs3s9tuuy16PY03j/vk2d5jvmbNmk1VtW6YZVcteutDSvIJ4AZzzHr2Ijazf1Wdn+SmwKeSfKOqvj97oapaD6wHWLduXU1PTy+63l6vx/asp/HmcZ88y3HMlyxYq+re881L8pMkN6yqC5LcELhwnm2c3/08O8lG4HbAHwSrJO1MRtXHegrwN93w3wAfnL1Akusk2b0bXg3cFfjWslUoSdtpVMH6MuA+Sb4H3KcbJ8m6JP/aLXMr4LQk/w18mn4fq8Eqaae3ZF0BC6mqnwH3mmP6acBjuuEvAn+8zKVJ0g7zk1eS1JjBKkmNGayS1JjBKkmNGayS1NjEB2sSkrBp06atw5K0IyY6WOcLUcNV0o6Y6GCVpKVgsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYwarJDVmsEpSYxMdrFW1qOmSNIxVoy5g1GZCtNfrGaiSmpjoFqskLQWDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqTGDVZIaM1glqbGstG8mTfJToLcdq64GLmpcjnZ+HvfJs73HfLqq9hlmwRUXrNsryWlVtW7UdWh5edwnz3Icc7sCJKkxg1WSGjNYf2/9qAvQSHjcJ8+SH3P7WCWpMVusktSYwSpJjU1MsCa5fNQ1qI0kleQdA+Orkvw0yYcXuZ2NSdZ1w6cm2bt1rVo+s//GkzwyyRu64b9L8ohtrL91+R21qsVGpGX2S+CgJHtU1a+A+wDn7cgGq+p+TSrTTqmq3rSc+5uYFutckkwn+WSSM7qf+yfZJcnZ6ds7ye+S3K1b/nNJbjbqugXAR4D7d8NHAifNzEiyZ5ITknw1ydeTHNZN3yPJe7rj/V5gj4F1zkmyOsmaJN8cmP70JM/vhjcmeU2Szyb5dpKDk7w/yfeS/OMyPGdtpyTPT/L0bvjg7nfgS0leOXi8gRsl+c/umL5ie/c30cEKvAF4e1XdGngX8Lqq2gJ8FzgQ+BNgE/CnSXZ286R6AAADu0lEQVQHblxVZ42sWg16D3BEkqsDtwa+MjDv2cCnqupg4B7AK5PsCTwOuKI73i8G7rAd+91cVXcD3gR8EHg8cBDwyCTX2+5noxb2SHL6zAN44TzLvRX4u6q6C7Bl1rzbAg8F/hh4aJL9tqeQSQ/WuwDv7obfQT9IAT4H3K17vLSbfjDw1eUuUHOrqjOANfRbq6fOmv1nwLHdH9dG4OrA/vSP5zsH1j9jO3Z9SvfzG8CZVXVBVf0GOBvYrj9CNfOrqrrtzAN47uwFun70varqi92kd89a5JNVdUlV/Rr4FjC9PYVMerDONnNR7+eAPwXuSP+Pdm/gEOCzoylL8zgFeBUD3QCdAIcP/JHtX1Xf7uZt68LtK7nq38XVZ83/TffzdwPDM+Oes9j5ZRvzB4/pFrbzmE56sH4ROKIbPgr4fDf8FeB/Ab/r/nOdDvwt/cDVzuME4IVV9Y1Z0z8K/N8kAUhyu276Z+kfZ5IcRL8LYbafANdPcr2u++cBS1K5RqKqfgFcluTO3aQjFlp+e01SsF4jybkDj6cCTwQeleQM4K+BJwF0b+1+BHy5W/dzwF703/5pJ1FV51bVa+eY9SJgV+CM7sTEi7rpxwHX7I73M4D/mmObv6XfN/cV4MPA/yxF7RqpRwPrk3yJfgv2ktY78COtkiZKkmtW1eXd8LHADavqSS33YZ+QpElz/yTPpJ9/PeCRrXdgi1WSGpukPlZJWhYGqyQ1ZrBKUmMGq1aUJDfo7gfw/STf6u5adfNZnweXlpRXBWjF6D4Q8O/AiVV1RDfttsC+Iy1ME8cWq1aSewC/HbxFXFWdTv/DHgB0d6/6XJKvdY//1U2/YXfXqtOTfDPJn3Z3OntbN/6NJE9Z/qekcWSLVSvJQfTvRraQC4H7VNWvkxxA/z4D64CHAR+tqhcn2QW4Bv07HU1V1UGw9QYe0jYZrJo0uwJv6LoItgA376Z/FTghya7AB6rq9CRnAzdN8nrgP4CPjaRijR27ArSSnMm277H6FPo3WrkN/ZbqbgBV9Vn6txU8D3hHkkd0N+y4Df1bDz4e+NelKVsrjcGqleRTwO5JHjszIcnBXPWemtcGLqiq39G/8c4u3XLTwIVV9RbgeOD2SVYDV6uqDcBzgNsvz9PQuLMrQCtGVVWSvwD+ubu5xq+Bc4AnDyz2RmBDkocAn6b//VnQv9/u3yf5LXA58AhgCnhrkpkGyDOX/EloRfBeAZLUmF0BktSYwSpJjRmsktSYwSpJjRmsktSYwSpJjRmsktTY/wcH+tt3Cplc7gAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 360x720 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "full_data = np.column_stack((perturbed_samples,\n",
    "                             class_perturb_samples))\n",
    "ld.lime_display(full_data, lime_beta, lime_int, x, x_class, features,\n",
    "                ['Low', 'Medium', 'High'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}