cnw_lime_plots.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "0c38c553",
   "metadata": {},
   "source": [
    "# LIME plots\n",
    "\n",
    "Ressources:\n",
    "\n",
    "https://marcotcr.github.io/lime/tutorials/Tutorial%20-%20continuous%20and%20categorical%20features.html\n",
    "\n",
    "https://medium.com/mlearning-ai/shap-force-plots-for-classification-d30be430e195\n",
    "https://stackoverflow.com/questions/65549588/shap-treeexplainer-for-randomforest-multiclass-what-is-shap-valuesi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "57cce01f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.tree import DecisionTreeClassifier\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from xgboost import XGBClassifier\n",
    "\n",
    "from evaluation import evaluation_classification\n",
    "\n",
    "import lime\n",
    "import lime.lime_tabular\n",
    "import numpy as np\n",
    "import random\n",
    "import shap\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "import sklearn\n",
    "import seaborn as sn\n",
    "from joblib import load\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "import matplotlib"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "48c2e766",
   "metadata": {},
   "outputs": [],
   "source": [
    "# read all prepared data\n",
    "\n",
    "name_array = [\"X_train_sub\", \"y_train_sub\", \"X_exp_train_sub\", \"y_exp_train_sub\", \\\n",
    "              \"X_train\", \"y_train\", \"X_test\", \"y_test\", \\\n",
    "              \"X_exp_train\", \"y_exp_train\", \"X_exp_test\", \"y_exp_test\", \\\n",
    "              \"X_test_urban\", \"y_test_urban\", \"X_test_rural\", \"y_test_rural\", \\\n",
    "              \"X_test_exp_urban\", \"y_test_exp_urban\", \"X_test_exp_rural\", \"y_test_exp_rural\", \\\n",
    "              \"X_train_exp_urban\", \"y_train_exp_urban\", \"X_train_exp_rural\", \"y_train_exp_rural\"]\n",
    "name_data = []\n",
    "\n",
    "for i in range(len(name_array)):\n",
    "    name_data.append(np.load(\"mobility_data/network_new_split_shuffled/\" + name_array[i] + \".npy\", allow_pickle=True))\n",
    "\n",
    "X_train_sub, y_train_sub, X_exp_train_sub, y_exp_train_sub, \\\n",
    "X_train, y_train, X_test, y_test, \\\n",
    "X_exp_train, y_exp_train, X_exp_test, y_exp_test, \\\n",
    "X_test_urban, y_test_urban, X_test_rural, y_test_rural, \\\n",
    "X_test_exp_urban, y_test_exp_urban, X_test_exp_rural, y_test_exp_rural, \\\n",
    "X_train_exp_urban, y_train_exp_urban, X_train_exp_rural, y_train_exp_rural = \\\n",
    "name_data[0], name_data[1], name_data[2], name_data[3], \\\n",
    "name_data[4], name_data[5], name_data[6], name_data[7], \\\n",
    "name_data[8], name_data[9], name_data[10], name_data[11], \\\n",
    "name_data[12], name_data[13], name_data[14], name_data[15], \\\n",
    "name_data[16], name_data[17], name_data[18], name_data[19], \\\n",
    "name_data[20], name_data[21], name_data[22], name_data[23]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b4ea8a18",
   "metadata": {},
   "outputs": [],
   "source": [
    "# load models\n",
    "\n",
    "dt = load(\"models_shuffled/decision_tree.joblib\")\n",
    "rf = load(\"models_shuffled/random_forest.joblib\")\n",
    "xgb = load(\"models_shuffled/xgboost.joblib\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c1cee29d",
   "metadata": {},
   "source": [
    "# EXPLANATIONS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "77dc6f64",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "randomly generated indices [98, 107, 10, 66, 130, 124, 103, 77, 122, 91]\n",
      "original label [4, 9, 9, 9, 5, 9, 8, 9, 9, 3]\n",
      "prediction rf [2, 9, 9, 9, 7, 9, 7, 9, 9, 1]\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n",
      "np.matrix usage is deprecated in 1.0 and will raise a TypeError in 1.2. Please convert to a numpy array with np.asarray. For more information see: https://numpy.org/doc/stable/reference/generated/numpy.matrix.html\n"
     ]
    }
   ],
   "source": [
    "# local explanation parameters\n",
    "\n",
    "# N - number of instances to look at\n",
    "N = 10\n",
    "# indices\n",
    "random.seed(0)\n",
    "collect_i = random.sample(range(0, X_exp_test.shape[0]), N)\n",
    "\n",
    "orig_collect_p = []\n",
    "collect_p = []\n",
    "\n",
    "for c in range(N):\n",
    "    pred_i = rf.predict(np.asmatrix(X_exp_test[collect_i[c]]))[0]\n",
    "    collect_p.append(pred_i)\n",
    "    orig_collect_p.append(y_exp_test[collect_i[c]])\n",
    "\n",
    "print(\"randomly generated indices\", collect_i)\n",
    "print(\"original label\", orig_collect_p)\n",
    "print(\"prediction rf\", collect_p)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "2eabb296",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[9 1 7 5 2 2 0 9 3 9 9 9 9 6 6 1 1 9 2 1 2 9 6 6 6 1 5 1 9 0 1 0 1 1 2 1 9\n",
      " 9 9 9 7 8 9 5 1 8 9 1 1 3 8 0 1 9 2 1 9 7 9 7 8 1 1 9 8 7 9 8 6 5 5 9 2 0\n",
      " 9 9 1 9 8 9 9 2 9 0 0 0 6 9 0 1 2 1 7 2 1 9 1 5 2 1 8 9 9 7 0 1 3 9 1 7 0\n",
      " 0 9 8 0 9 9 9 1 0 1 9 9 8 9 9 0 3 9 6 7 3 9 8 9 7 1 9 6 0 4 9 9 8]\n",
      "144 60 84\n"
     ]
    }
   ],
   "source": [
    "N_all = len(X_exp_test)\n",
    "prediction = rf.predict(X_exp_test)\n",
    "print(rf.predict(X_exp_test))\n",
    "\n",
    "N_all_urban = len(X_test_exp_urban)\n",
    "N_all_rural = len(X_test_exp_rural)\n",
    "\n",
    "prediction_urban = rf.predict(X_test_exp_urban)\n",
    "prediction_rural = rf.predict(X_test_exp_rural)\n",
    "\n",
    "print(N_all, N_all_urban, N_all_rural)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "5c1010ef",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(144, 18)\n",
      "(333, 18)\n"
     ]
    }
   ],
   "source": [
    "print(np.shape(X_exp_test))\n",
    "print(np.shape(X_exp_train))\n",
    "\n",
    "columns = [\"te_cn\", \"out_cn\", \"in_cn\", \"dc_cn\", \"cc_cn\", \"out/in_cn\", \"te_cl\", \"out_cl\", \"in_cl\", \"dc_cl\", \"cc_cl\", \"out/in_cl\", \"te_sn\", \"out_sn\", \"in_sn\", \"dc_sn\", \"cc_sn\", \"out/in_sn\"]\n",
    "target = [\"0\", \"1\", \"2\", \"3\", \"4\", \"5\", \"6\", \"7\", \"8\", \"9\"]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4c571e7c",
   "metadata": {},
   "source": [
    "# 1) LIME"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "c90fa249",
   "metadata": {},
   "outputs": [],
   "source": [
    "values_ = np.load(\"explanations/lime_matrix.npy\")\n",
    "# training on urban or rural training set, N adapted\n",
    "values_urban_2 = np.load(\"explanations_refined/lime_matrix_urban.npy\")\n",
    "values_rural_2 = np.load(\"explanations_refined/lime_matrix_rural.npy\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "4ac53f27",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(10, 18, 144)\n"
     ]
    }
   ],
   "source": [
    "print(np.shape(values_))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "1a9ae4af",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# filter only non-zero\n",
    "\n",
    "plt.figure(figsize = (10, 6))\n",
    "#plt.imshow(np.sum(values_, axis = 2) / 144,cmap = \"Reds\")\n",
    "# new scale\n",
    "plt.imshow(np.sum(values_, axis = 2) / 144,cmap = \"Reds\", vmin = -0.01, vmax = 0.06)\n",
    "#plt.imshow(np.sum(values_, axis = 2) / np.count_nonzero(values_, axis = 2), cmap = \"coolwarm\")\n",
    "plt.colorbar(shrink = 0.75)\n",
    "l = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]\n",
    "c = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n",
    "plt.xticks(l, columns, rotation='vertical')\n",
    "plt.yticks(c)\n",
    "plt.ylabel(\"predicted class\")\n",
    "plt.tight_layout()\n",
    "\n",
    "plt.savefig(\"explanations_new_colorbar/lime_matrix.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "5c3e686c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# filter only non-zero\n",
    "\n",
    "plt.figure(figsize = (10, 6))\n",
    "plt.imshow((np.sum(values_urban_2, axis = 2) / 60), cmap = \"Reds\")\n",
    "#plt.imshow(np.sum(values_, axis = 2) / np.count_nonzero(values_, axis = 2), cmap = \"coolwarm\")\n",
    "plt.colorbar(shrink = 0.75)\n",
    "l = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]\n",
    "c = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n",
    "plt.xticks(l, columns, rotation='vertical')\n",
    "plt.yticks(c)\n",
    "plt.ylabel(\"predicted class\")\n",
    "plt.tight_layout()\n",
    "\n",
    "\n",
    "plt.savefig(\"explanations_new_colorbar/lime_matrix_urban_2.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "1d54312b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x432 with 2 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# filter only non-zero\n",
    "\n",
    "plt.figure(figsize = (10, 6))\n",
    "plt.imshow(np.sum(values_rural_2, axis = 2) / 84, cmap = \"Reds\")\n",
    "#plt.imshow(np.sum(values_, axis = 2) / np.count_nonzero(values_, axis = 2), cmap = \"coolwarm\")\n",
    "plt.colorbar(shrink = 0.75)\n",
    "l = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17]\n",
    "c = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9]\n",
    "plt.xticks(l, columns, rotation='vertical')\n",
    "plt.yticks(c)\n",
    "plt.ylabel(\"predicted class\")\n",
    "plt.tight_layout()\n",
    "\n",
    "\n",
    "plt.savefig(\"explanations_new_colorbar/lime_matrix_rural_2.pdf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b7f85a63",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "late",
   "language": "python",
   "name": "late"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}