VPN_5.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "56db7f42",
   "metadata": {},
   "source": [
    "## Prepare Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72e2646e-796d-4f29-b203-7df69e24fe7a",
   "metadata": {},
   "source": [
    "### Flow count for VPN and non-VPN"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "771f0e5f-bbf4-4226-86c5-d4ac170fd04a",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "label\n",
      "NoVPN    43597\n",
      "VPN       8350\n",
      "Name: count, dtype: int64\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 800x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "df = pd.read_csv('./VNAT/WF/flows_extra_labelled_5.csv',low_memory=False)\n",
    "\n",
    "label_counts = df['label'].value_counts()\n",
    "print(label_counts)\n",
    "\n",
    "plt.figure(figsize=(8, 6))  \n",
    "label_counts.plot(kind='bar', color=['skyblue', 'red'])  \n",
    "\n",
    "plt.title('Count of Labels in CSV', fontsize=16)\n",
    "plt.xlabel('Label', fontsize=14)\n",
    "plt.ylabel('Count', fontsize=14)\n",
    "\n",
    "plt.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f6854cdf-2214-487f-bb14-ed27132678a4",
   "metadata": {},
   "source": [
    "### Create a dataset for Wavelet Features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3d29e0a5-e2a3-4794-8ec4-27b03d525153",
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "df1 = pd.read_csv('./VNAT/WF/flows_extra_labelled_5.csv',low_memory=False)\n",
    "\n",
    "columns_to_select = [\n",
    "                     #select Wavelet features\n",
    "                     #forward\n",
    "                    'udps.f_p_k_0','udps.f_p_k_1','udps.f_p_k_2','udps.f_p_k_3','udps.f_p_k_4','udps.f_p_k_5',\n",
    "                     'udps.f_u_k_0','udps.f_u_k_1','udps.f_u_k_2','udps.f_u_k_3','udps.f_u_k_4','udps.f_u_k_5',\n",
    "                     'udps.f_sigma_k_0','udps.f_sigma_k_1','udps.f_sigma_k_2','udps.f_sigma_k_3','udps.f_sigma_k_4','udps.f_sigma_k_5',\n",
    "                     'udps.f_S_k_0','udps.f_S_k_1','udps.f_S_k_2','udps.f_S_k_3','udps.f_S_k_4','udps.f_S_k_5',\n",
    "                  #backward\n",
    "                     'udps.b_p_k_0','udps.b_p_k_1','udps.b_p_k_2','udps.b_p_k_3','udps.b_p_k_4','udps.b_p_k_5',\n",
    "                     'udps.b_u_k_0','udps.b_u_k_1','udps.b_u_k_2','udps.b_u_k_3','udps.b_u_k_4','udps.b_u_k_5',\n",
    "                     'udps.b_sigma_k_0','udps.b_sigma_k_1','udps.b_sigma_k_2','udps.b_sigma_k_3','udps.b_sigma_k_4','udps.b_sigma_k_5',\n",
    "                     'udps.b_S_k_0','udps.b_S_k_1','udps.b_S_k_2','udps.b_S_k_3','udps.b_S_k_4','udps.b_S_k_5'\n",
    "                   , 'label','category']\n",
    "df_WNF = df1[columns_to_select]                  \n",
    "\n",
    "df_WNF.to_csv('./VNAT/WF/flows_extra_labelled_WF5.csv', index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e14e4aad-250e-44d4-b5f4-3bfa61ac91c4",
   "metadata": {},
   "source": [
    "### Rndom Forest"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "47f658ab-3b9a-4b40-a9d9-c84458990873",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Accuracy: 0.999975936665303\n",
      "Test Accuracy: 0.9910490856592877\n",
      "\n",
      "Confusion Matrix:\n",
      "[[0.9952987  0.0047013 ]\n",
      " [0.03115638 0.96884362]]\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "       NoVPN       0.99      1.00      0.99      8721\n",
      "         VPN       0.98      0.97      0.97      1669\n",
      "\n",
      "    accuracy                           0.99     10390\n",
      "   macro avg       0.98      0.98      0.98     10390\n",
      "weighted avg       0.99      0.99      0.99     10390\n",
      "\n",
      "\n",
      "Detailed Misclassification Summary (per Category):\n",
      "            category true_label predicted_label  count\n",
      "0               Chat      NoVPN             VPN     23\n",
      "1               Chat        VPN           NoVPN     33\n",
      "2  Command & Control      NoVPN             VPN      4\n",
      "3  Command & Control        VPN           NoVPN      5\n",
      "4      File Transfer      NoVPN             VPN      3\n",
      "5      File Transfer        VPN           NoVPN     10\n",
      "6          Streaming      NoVPN             VPN     11\n",
      "7          Streaming        VPN           NoVPN      4\n",
      "Confusion matrix saved at: output_results\\confusion_matrix_RF5.png\n",
      "Accuracy and classification report saved at: output_results\\RF5.txt\n",
      "Detailed misclassification summary saved at: output_results\\RF5.csv\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.ensemble import RandomForestClassifier\n",
    "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
    "\n",
    "output_dir = \"output_results\"\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "\n",
    "file_path = './VNAT/WF/flows_extra_labelled_WF5.csv'  \n",
    "df = pd.read_csv(file_path)\n",
    "\n",
    "df['label_encoded'] = df['label'].map({'NoVPN': 0, 'VPN': 1})\n",
    "\n",
    "category = df['category']\n",
    "\n",
    "X = df.drop(['label', 'label_encoded', 'category'], axis=1)\n",
    "y = df['label_encoded']\n",
    "\n",
    "X_train, X_test, y_train, y_test, category_train, category_test = train_test_split(\n",
    "    X, y, category, test_size=0.2, random_state=42\n",
    ")\n",
    "\n",
    "clf = RandomForestClassifier(random_state=42)\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "train_accuracy = accuracy_score(y_train, clf.predict(X_train))\n",
    "test_accuracy = accuracy_score(y_test, y_pred)\n",
    "\n",
    "cm = confusion_matrix(y_test, y_pred, normalize='true')\n",
    "\n",
    "print(\"Train Accuracy:\", train_accuracy)\n",
    "print(\"Test Accuracy:\", test_accuracy)\n",
    "print(\"\\nConfusion Matrix:\")\n",
    "print(cm)\n",
    "print(\"\\nClassification Report:\")\n",
    "print(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "plt.figure(figsize=(6, 5))\n",
    "sns.heatmap(cm, annot=True, fmt=\".2f\", cmap=\"Blues\", xticklabels=['non-VPN', 'VPN'], \n",
    "yticklabels=['non-VPN', 'VPN'], annot_kws={\"size\": 16})  \n",
    "\n",
    "plt.xlabel(\"Predicted Label\", fontsize=20)\n",
    "plt.ylabel(\"True Label\", fontsize=20)\n",
    "plt.title(\"\", fontsize=20)\n",
    "plt.xticks(fontsize=20)\n",
    "plt.yticks(fontsize=20)\n",
    "plt.tight_layout()\n",
    "\n",
    "conf_matrix_path = os.path.join(output_dir, \"confusion_matrix_RF5.png\")\n",
    "plt.savefig(conf_matrix_path, dpi=300, bbox_inches='tight')\n",
    "plt.close()\n",
    "\n",
    "accuracy_report_path = os.path.join(output_dir, \"RF5.txt\")\n",
    "with open(accuracy_report_path, \"w\") as f:\n",
    "    f.write(f\"Train Accuracy: {train_accuracy:.4f}\\n\")\n",
    "    f.write(f\"Test Accuracy: {test_accuracy:.4f}\\n\\n\")\n",
    "    f.write(\"Classification Report:\\n\")\n",
    "    f.write(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "misclassified_mask = (y_test != y_pred)\n",
    "misclassified_samples = X_test.loc[misclassified_mask].copy()\n",
    "\n",
    "misclassified_samples['true_label'] = y_test.loc[misclassified_mask]\n",
    "misclassified_samples['predicted_label'] = y_pred[misclassified_mask]\n",
    "misclassified_samples['category'] = category_test.loc[misclassified_mask]\n",
    "\n",
    "label_map = {0: 'NoVPN', 1: 'VPN'}\n",
    "misclassified_samples['true_label'] = misclassified_samples['true_label'].map(label_map)\n",
    "misclassified_samples['predicted_label'] = misclassified_samples['predicted_label'].map(label_map)\n",
    "\n",
    "misclassification_detail = misclassified_samples.groupby(['category', 'true_label', 'predicted_label']).size().reset_index(name='count')\n",
    "\n",
    "print(\"\\nDetailed Misclassification Summary (per Category):\")\n",
    "print(misclassification_detail)\n",
    "\n",
    "misclassification_detail_path = os.path.join(output_dir, \"RF5.csv\")\n",
    "misclassification_detail.to_csv(misclassification_detail_path, index=False)\n",
    "\n",
    "print(\"Confusion matrix saved at:\", conf_matrix_path)\n",
    "print(\"Accuracy and classification report saved at:\", accuracy_report_path)\n",
    "print(\"Detailed misclassification summary saved at:\", misclassification_detail_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "024cf7b6-2531-49e4-865c-fee06c3051db",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA94AAAKACAYAAABqnqAYAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjkuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8hTgPZAAAACXBIWXMAAA9hAAAPYQGoP6dpAAB5nUlEQVR4nO3deXxTVf7/8Xfa0hYoLavUQqGAqCBIlR0VVCqoKIuOC84IMjKOfhVF3AAVRnTElR1lHB0dnUFRB5RRQRBQRHBhURTZhEJRLItCQZBFcn5/8GuG0EKT9tzkNLyej0ce2pubTz/nk5Nz8+Hmpj5jjBEAAAAAAPBEXLQTAAAAAAAgltF4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwBQzr300kvy+XzasGFDtFMBAADFoPEGAJQ7hY1mcbfBgwd78jsXLlyov/zlL9q5c6cn8U9ke/fu1V/+8hd9+OGH0U4FAABPJEQ7AQAASmvEiBFq0KBB0LZmzZp58rsWLlyohx56SDfccIOqVq3qye8oreuvv17XXnutkpKSop1Kqezdu1cPPfSQJOn888+PbjIAAHiAxhsAUG5dcsklatWqVbTTKJM9e/aocuXKZYoRHx+v+Ph4SxlFjt/v14EDB6KdBgAAnuOj5gCAmDVjxgydd955qly5sqpUqaJu3bppxYoVQfssX75cN9xwgxo2bKjk5GSlp6frj3/8o3766afAPn/5y190zz33SJIaNGgQ+Fj7hg0btGHDBvl8Pr300ktFfr/P59Nf/vKXoDg+n0/ffvutrrvuOlWrVk3nnntu4P5//etfatmypSpWrKjq1avr2muv1aZNm0ocZ3HXeGdlZemyyy7Thx9+qFatWqlixYpq3rx54OPcU6dOVfPmzZWcnKyWLVtq2bJlQTFvuOEGpaSkaP369eratasqV66sjIwMjRgxQsaYoH337Nmju+66S5mZmUpKStJpp52mp556qsh+Pp9Pt912m/7973/rjDPOUFJSkiZNmqRatWpJkh566KFAbQvrFsrzc2Rtv/vuu8CnEtLS0tSvXz/t3bu3SM3+9a9/qU2bNqpUqZKqVaumjh07atasWUH7hDJ/8vPz1a9fP9WtW1dJSUk6+eST1aNHD663BwAE4Yw3AKDcKigo0Pbt24O21axZU5L0yiuvqG/fvuratasef/xx7d27V88++6zOPfdcLVu2TFlZWZKk2bNna/369erXr5/S09O1YsUKPffcc1qxYoU+/fRT+Xw+XXHFFVqzZo1effVVjR49OvA7atWqpW3btoWd91VXXaXGjRvr0UcfDTSnf/3rX/Xggw/q6quvVv/+/bVt2zaNHz9eHTt21LJly0r18fbvvvtO1113nf785z/rD3/4g5566ildfvnlmjRpkoYOHar/+7//kySNHDlSV199tVavXq24uP/9m/yhQ4d08cUXq127dnriiSc0c+ZMDR8+XL/99ptGjBghSTLGqHv37po3b55uvPFGZWdn6/3339c999yjH374QaNHjw7Kae7cuXr99dd12223qWbNmmrRooWeffZZ3XLLLerVq5euuOIKSdKZZ54pKbTn50hXX321GjRooJEjR2rp0qV6/vnnddJJJ+nxxx8P7PPQQw/pL3/5izp06KARI0YoMTFRn332mebOnasuXbpICn3+XHnllVqxYoUGDBigrKwsbd26VbNnz1ZeXl5gHwAAZAAAKGdefPFFI6nYmzHG7N6921StWtX86U9/Cnpcfn6+SUtLC9q+d+/eIvFfffVVI8nMnz8/sO3JJ580kkxubm7Qvrm5uUaSefHFF4vEkWSGDx8e+Hn48OFGkundu3fQfhs2bDDx8fHmr3/9a9D2r7/+2iQkJBTZfqx6HJlb/fr1jSSzcOHCwLb333/fSDIVK1Y0GzduDGz/29/+ZiSZefPmBbb17dvXSDIDBgwIbPP7/aZbt24mMTHRbNu2zRhjzFtvvWUkmUceeSQop9/97nfG5/OZ7777LqgecXFxZsWKFUH7btu2rUitCoX6/BTW9o9//GPQvr169TI1atQI/Lx27VoTFxdnevXqZQ4dOhS0r9/vN8aEPn927NhhJJknn3yySI4AAByJj5oDAMqtiRMnavbs2UE36fBZ0p07d6p3797avn174BYfH6+2bdtq3rx5gRgVK1YM/P++ffu0fft2tWvXTpK0dOlST/K++eabg36eOnWq/H6/rr766qB809PT1bhx46B8w9G0aVO1b98+8HPbtm0lSRdeeKHq1atXZPv69euLxLjtttsC/1/4UfEDBw7ogw8+kCS99957io+P1+233x70uLvuukvGGM2YMSNoe6dOndS0adOQxxDu83N0bc877zz99NNP2rVrlyTprbfekt/v17Bhw4LO7heOTwp9/lSsWFGJiYn68MMPtWPHjpDHBAA48fBRcwBAudWmTZtiv1xt7dq1kg43mMVJTU0N/P/PP/+shx56SK+99pq2bt0atF9BQYHFbP/n6G9iX7t2rYwxaty4cbH7V6hQoVS/58jmWpLS0tIkSZmZmcVuP7p5jIuLU8OGDYO2nXrqqZIUuIZ548aNysjIUJUqVYL2a9KkSeD+Ix099pKE+/wcPeZq1apJOjy21NRUrVu3TnFxccdt/kOdP0lJSXr88cd11113qXbt2mrXrp0uu+wy9enTR+np6aEPEgAQ82i8AQAxx+/3Szp8nW5xDVBCwv8Of1dffbUWLlyoe+65R9nZ2UpJSZHf79fFF18ciHM8R19jXOjQoUPHfMyRZ3EL8/X5fJoxY0ax306ekpJSYh7FOdY3nR9ruznqy9C8cPTYSxLu82NjbOHMn4EDB+ryyy/XW2+9pffff18PPvigRo4cqblz5+qss84K+XcCAGIbjTcAIOY0atRIknTSSScpJyfnmPvt2LFDc+bM0UMPPaRhw4YFthee8TzSsRrswjOqO3fuDNp+9JnekvI1xqhBgwaBM8ou8Pv9Wr9+fVBOa9askaTAF4fVr19fH3zwgXbv3h101nvVqlWB+0tyrNqG8/yEqlGjRvL7/fr222+VnZ19zH2kkufPkfvfdddduuuuu7R27VplZ2fr6aef1r/+9a9S5wkAiC1c4w0AiDldu3ZVamqqHn30UR08eLDI/YXfRF54dvTos6Fjxowp8pjCv7V9dIOdmpqqmjVrav78+UHbn3nmmZDzveKKKxQfH6+HHnqoSC7GmCJ/OiuSJkyYEJTLhAkTVKFCBXXu3FmSdOmll+rQoUNB+0nS6NGj5fP5dMkll5T4OypVqiSpaG3DeX5C1bNnT8XFxWnEiBFFzpgX/p5Q58/evXu1b9++oPsaNWqkKlWqaP/+/aXOEQAQezjjDQCIOampqXr22Wd1/fXX6+yzz9a1116rWrVqKS8vT++++67OOeccTZgwQampqerYsaOeeOIJHTx4UHXq1NGsWbOUm5tbJGbLli0lSffff7+uvfZaVahQQZdffrkqV66s/v3767HHHlP//v3VqlUrzZ8/P3BmOBSNGjXSI488oiFDhmjDhg3q2bOnqlSpotzcXE2bNk033XST7r77bmv1CVVycrJmzpypvn37qm3btpoxY4beffddDR06NPC3ty+//HJdcMEFuv/++7Vhwwa1aNFCs2bN0ttvv62BAwcGzh4fT8WKFdW0aVNNmTJFp556qqpXr65mzZqpWbNmIT8/oTrllFN0//336+GHH9Z5552nK664QklJSfriiy+UkZGhkSNHhjx/1qxZo86dO+vqq69W06ZNlZCQoGnTpmnLli269tprS50jACD20HgDAGLSddddp4yMDD322GN68skntX//ftWpU0fnnXee+vXrF9hv8uTJGjBggCZOnChjjLp06aIZM2YoIyMjKF7r1q318MMPa9KkSZo5c6b8fr9yc3NVuXJlDRs2TNu2bdObb76p119/XZdccolmzJihk046KeR8Bw8erFNPPVWjR4/WQw89JOnwl6B16dJF3bt3t1OUMMXHx2vmzJm65ZZbdM8996hKlSoaPnx40Me+4+LiNH36dA0bNkxTpkzRiy++qKysLD355JO66667Qv5dzz//vAYMGKA777xTBw4c0PDhw9WsWbOQn59wjBgxQg0aNND48eN1//33q1KlSjrzzDN1/fXXB/YJZf5kZmaqd+/emjNnjl555RUlJCTo9NNP1+uvv64rr7yy1PkBAGKPz0Tim1QAAEC5csMNN+jNN9/UL7/8Eu1UAAAo97jGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8xDXeAAAAAAB4iDPeAAAAAAB4KCb+nJjf79fmzZtVpUoV+Xy+aKcDAAAAAIhxxhjt3r1bGRkZios7/jntmGi8N2/erMzMzGinAQAAAAA4wWzatEl169Y97j6larwnTpyoJ598Uvn5+WrRooXGjx+vNm3aFLvvihUrNGzYMC1ZskQbN27U6NGjNXDgwKB9Ro4cqalTp2rVqlWqWLGiOnTooMcff1ynnXZaSPlUqVJF0uEBp6amlmZIAAAAAACEbNeuXcrMzAz0o8cTduM9ZcoUDRo0SJMmTVLbtm01ZswYde3aVatXr9ZJJ51UZP+9e/eqYcOGuuqqq3TnnXcWG/Ojjz7SrbfeqtatW+u3337T0KFD1aVLF3377beqXLlyiTkVfrw8NTWVxhsAAAAAEDGhXO4c9reat23bVq1bt9aECRMkHb6+OjMzUwMGDNDgwYOP+9isrCwNHDiwyBnvo23btk0nnXSSPvroI3Xs2LHI/fv379f+/fsDPxf+S0NBQQGNNwAAAADAc7t27VJaWlpIfWhY32p+4MABLVmyRDk5Of8LEBennJwcLVq0qHTZFqOgoECSVL169WLvHzlypNLS0gI3ru8GAAAAALgqrMZ7+/btOnTokGrXrh20vXbt2srPz7eSkN/v18CBA3XOOeeoWbNmxe4zZMgQFRQUBG6bNm2y8rsBAAAAALDNuW81v/XWW/XNN99owYIFx9wnKSlJSUlJEcwKAAAAAIDSCavxrlmzpuLj47Vly5ag7Vu2bFF6enqZk7ntttv0zjvvaP78+SV+HTsAAAAAAOVBWB81T0xMVMuWLTVnzpzANr/frzlz5qh9+/alTsIYo9tuu03Tpk3T3Llz1aBBg1LHAgAAAADAJWF/1HzQoEHq27evWrVqpTZt2mjMmDHas2eP+vXrJ0nq06eP6tSpo5EjR0o6/IVs3377beD/f/jhB3355ZdKSUnRKaecIunwx8snT56st99+W1WqVAlcL56WlqaKFStaGSgAAAAAANEQ9p8Tk6QJEyboySefVH5+vrKzszVu3Di1bdtWknT++ecrKytLL730kiRpw4YNxZ7B7tSpkz788MPDSRzj7569+OKLuuGGG0rMJ5yvcQcAAAAAoKzC6UNL1Xi7hsYbAAAAABBJnv0dbwAAAAAAEB4abwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPJQQ7QSixucr/WONsZcHAAAAACCmccYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHipV4z1x4kRlZWUpOTlZbdu21eeff37MfVesWKErr7xSWVlZ8vl8GjNmTJljAgAAAABQXoTdeE+ZMkWDBg3S8OHDtXTpUrVo0UJdu3bV1q1bi91/7969atiwoR577DGlp6dbiQkAAAAAQHnhM8aYcB7Qtm1btW7dWhMmTJAk+f1+ZWZmasCAARo8ePBxH5uVlaWBAwdq4MCB1mJK0q5du5SWlqaCggKlpqaGNhCfL7T9ihNeyQAAAAAAMSacPjSsM94HDhzQkiVLlJOT878AcXHKycnRokWLSpVsaWLu379fu3btCroBAAAAAOCisBrv7du369ChQ6pdu3bQ9tq1ays/P79UCZQm5siRI5WWlha4ZWZmlup3AwAAAADgtXL5reZDhgxRQUFB4LZp06ZopwQAAAAAQLESwtm5Zs2aio+P15YtW4K2b9my5ZhfnOZFzKSkJCUlJZXq9wEAAAAAEElhnfFOTExUy5YtNWfOnMA2v9+vOXPmqH379qVKwIuYAAAAAAC4Iqwz3pI0aNAg9e3bV61atVKbNm00ZswY7dmzR/369ZMk9enTR3Xq1NHIkSMlHf7ytG+//Tbw/z/88IO+/PJLpaSk6JRTTgkpJgAAAAAA5VXYjfc111yjbdu2adiwYcrPz1d2drZmzpwZ+HK0vLw8xcX970T65s2bddZZZwV+fuqpp/TUU0+pU6dO+vDDD0OKCQAAAABAeRX23/F2EX/HGwAAAAAQSZ79HW8AAAAAABAeGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQwnRTqDc8/lK/1hj7OUBAAAAAHASZ7wBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB4qVeM9ceJEZWVlKTk5WW3bttXnn39+3P3feOMNnX766UpOTlbz5s313nvvBd3/yy+/6LbbblPdunVVsWJFNW3aVJMmTSpNagAAAAAAOCXsxnvKlCkaNGiQhg8frqVLl6pFixbq2rWrtm7dWuz+CxcuVO/evXXjjTdq2bJl6tmzp3r27KlvvvkmsM+gQYM0c+ZM/etf/9LKlSs1cOBA3XbbbZo+fXrpRwYAAAAAgAN8xhgTzgPatm2r1q1ba8KECZIkv9+vzMxMDRgwQIMHDy6y/zXXXKM9e/bonXfeCWxr166dsrOzA2e1mzVrpmuuuUYPPvhgYJ+WLVvqkksu0SOPPFJiTrt27VJaWpoKCgqUmpoa2kB8vtD2K86RJbMVBwAAAABQboTTh4Z1xvvAgQNasmSJcnJy/hcgLk45OTlatGhRsY9ZtGhR0P6S1LVr16D9O3TooOnTp+uHH36QMUbz5s3TmjVr1KVLl2Jj7t+/X7t27Qq6AQAAAADgorAa7+3bt+vQoUOqXbt20PbatWsrPz+/2Mfk5+eXuP/48ePVtGlT1a1bV4mJibr44os1ceJEdezYsdiYI0eOVFpaWuCWmZkZzjAAAAAAAIgYJ77VfPz48fr00081ffp0LVmyRE8//bRuvfVWffDBB8XuP2TIEBUUFARumzZtinDGAAAAAACEJiGcnWvWrKn4+Hht2bIlaPuWLVuUnp5e7GPS09OPu/+vv/6qoUOHatq0aerWrZsk6cwzz9SXX36pp556qsjH1CUpKSlJSUlJ4aQOAAAAAEBUhHXGOzExUS1bttScOXMC2/x+v+bMmaP27dsX+5j27dsH7S9Js2fPDux/8OBBHTx4UHFxwanEx8fL7/eHkx4AAAAAAM4J64y3dPhPf/Xt21etWrVSmzZtNGbMGO3Zs0f9+vWTJPXp00d16tTRyJEjJUl33HGHOnXqpKefflrdunXTa6+9psWLF+u5556TJKWmpqpTp0665557VLFiRdWvX18fffSRXn75ZY0aNcriUAEAAAAAiLywG+9rrrlG27Zt07Bhw5Sfn6/s7GzNnDkz8AVqeXl5QWevO3TooMmTJ+uBBx7Q0KFD1bhxY7311ltq1qxZYJ/XXntNQ4YM0e9//3v9/PPPql+/vv7617/q5ptvtjBEAAAAAACiJ+y/4+0i/o43AAAAACCSPPs73gAAAAAAIDw03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEMJ0U4AR/D5Sv9YY+zlAQAAAACwhjPeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQ6VqvCdOnKisrCwlJyerbdu2+vzzz4+7/xtvvKHTTz9dycnJat68ud57770i+6xcuVLdu3dXWlqaKleurNatWysvL6806QEAAAAA4IywG+8pU6Zo0KBBGj58uJYuXaoWLVqoa9eu2rp1a7H7L1y4UL1799aNN96oZcuWqWfPnurZs6e++eabwD7r1q3Tueeeq9NPP10ffvihli9frgcffFDJycmlHxkAAAAAAA7wGWNMOA9o27atWrdurQkTJkiS/H6/MjMzNWDAAA0ePLjI/tdcc4327Nmjd955J7CtXbt2ys7O1qRJkyRJ1157rSpUqKBXXnklpBz279+v/fv3B37etWuXMjMzVVBQoNTU1NAG4vOFtl9xjiyZrTi2YwEAAAAAPLNr1y6lpaWF1IeGdcb7wIEDWrJkiXJycv4XIC5OOTk5WrRoUbGPWbRoUdD+ktS1a9fA/n6/X++++65OPfVUde3aVSeddJLatm2rt95665h5jBw5UmlpaYFbZmZmOMMAAAAAACBiwmq8t2/frkOHDql27dpB22vXrq38/PxiH5Ofn3/c/bdu3apffvlFjz32mC6++GLNmjVLvXr10hVXXKGPPvqo2JhDhgxRQUFB4LZp06ZwhgEAAAAAQMQkRDsBv98vSerRo4fuvPNOSVJ2drYWLlyoSZMmqVOnTkUek5SUpKSkpIjmCQAAAABAaYR1xrtmzZqKj4/Xli1bgrZv2bJF6enpxT4mPT39uPvXrFlTCQkJatq0adA+TZo04VvNAQAAAADlXliNd2Jiolq2bKk5c+YEtvn9fs2ZM0ft27cv9jHt27cP2l+SZs+eHdg/MTFRrVu31urVq4P2WbNmjerXrx9OegAAAAAAOCfsj5oPGjRIffv2VatWrdSmTRuNGTNGe/bsUb9+/SRJffr0UZ06dTRy5EhJ0h133KFOnTrp6aefVrdu3fTaa69p8eLFeu655wIx77nnHl1zzTXq2LGjLrjgAs2cOVP//e9/9eGHH9oZJQAAAAAAURJ2433NNddo27ZtGjZsmPLz85Wdna2ZM2cGvkAtLy9PcXH/O5HeoUMHTZ48WQ888ICGDh2qxo0b66233lKzZs0C+/Tq1UuTJk3SyJEjdfvtt+u0007Tf/7zH5177rkWhggAAAAAQPSE/Xe8XRTO308L4O94AwAAAABKybO/4w0AAAAAAMJD4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAPJUQ7AXiAvwcOAAAAAM7gjDcAAAAAAB6i8QYAAAAAwEN81BzHx8fWAQAAAKBMOOMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIhvNUfk8A3pAAAAAE5AnPEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwUEK0EwDC5vOV/rHG2MsDAAAAAELAGW8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOChUjXeEydOVFZWlpKTk9W2bVt9/vnnx93/jTfe0Omnn67k5GQ1b95c77333jH3vfnmm+Xz+TRmzJjSpAYAAAAAgFPCbrynTJmiQYMGafjw4Vq6dKlatGihrl27auvWrcXuv3DhQvXu3Vs33nijli1bpp49e6pnz5765ptviuw7bdo0ffrpp8rIyAh/JAAAAAAAOCjsxnvUqFH605/+pH79+qlp06aaNGmSKlWqpH/84x/F7j927FhdfPHFuueee9SkSRM9/PDDOvvsszVhwoSg/X744QcNGDBA//73v1WhQoXSjQYAAAAAAMeE1XgfOHBAS5YsUU5Ozv8CxMUpJydHixYtKvYxixYtCtpfkrp27Rq0v9/v1/XXX6977rlHZ5xxRol57N+/X7t27Qq6AQAAAADgorAa7+3bt+vQoUOqXbt20PbatWsrPz+/2Mfk5+eXuP/jjz+uhIQE3X777SHlMXLkSKWlpQVumZmZ4QwDAAAAAICIifq3mi9ZskRjx47VSy+9JJ/PF9JjhgwZooKCgsBt06ZNHmcJAAAAAEDphNV416xZU/Hx8dqyZUvQ9i1btig9Pb3Yx6Snpx93/48//lhbt25VvXr1lJCQoISEBG3cuFF33XWXsrKyio2ZlJSk1NTUoBsAAAAAAC4Kq/FOTExUy5YtNWfOnMA2v9+vOXPmqH379sU+pn379kH7S9Ls2bMD+19//fVavny5vvzyy8AtIyND99xzj95///1wxwMAAAAAgFMSwn3AoEGD1LdvX7Vq1Upt2rTRmDFjtGfPHvXr10+S1KdPH9WpU0cjR46UJN1xxx3q1KmTnn76aXXr1k2vvfaaFi9erOeee06SVKNGDdWoUSPod1SoUEHp6ek67bTTyjo+AAAAAACiKuzG+5prrtG2bds0bNgw5efnKzs7WzNnzgx8gVpeXp7i4v53Ir1Dhw6aPHmyHnjgAQ0dOlSNGzfWW2+9pWbNmtkbBQAAAAAAjvIZY0y0kyirXbt2KS0tTQUFBaFf7x3iF7kV68iS2YpjM5aLOdmMZTMnAAAAACiFcPrQqH+rOQAAAAAAsYzGGwAAAAAAD4V9jTcQU/jYOgAAAACPccYbAAAAAAAPccYbsIEz5wAAAACOgTPeAAAAAAB4iDPegGs4ew4AAADEFM54AwAAAADgIRpvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB7iW82BWMY3pAMAAABRxxlvAAAAAAA8ROMNAAAAAICHaLwBAAAAAPAQjTcAAAAAAB7iy9UAlIwvaQMAAABKjTPeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADxE4w0AAAAAgIdovAEAAAAA8BCNNwAAAAAAHqLxBgAAAADAQzTeAAAAAAB4KCHaCQA4wfh8pX+sMfbyAAAAACKEM94AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPJQQ7QQAoFR8vtI/1hh7eQAAAAAl4Iw3AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwUKka74kTJyorK0vJyclq27atPv/88+Pu/8Ybb+j0009XcnKymjdvrvfeey9w38GDB3XfffepefPmqly5sjIyMtSnTx9t3ry5NKkBAAAAAOCUsBvvKVOmaNCgQRo+fLiWLl2qFi1aqGvXrtq6dWux+y9cuFC9e/fWjTfeqGXLlqlnz57q2bOnvvnmG0nS3r17tXTpUj344INaunSppk6dqtWrV6t79+5lGxkAAAAAAA7wGWNMOA9o27atWrdurQkTJkiS/H6/MjMzNWDAAA0ePLjI/tdcc4327Nmjd955J7CtXbt2ys7O1qRJk4r9HV988YXatGmjjRs3ql69eiXmtGvXLqWlpamgoECpqamhDcTnC22/4hxZMltxbMZyMSebscgp8rFiPScAAAAgTOH0oWGd8T5w4ICWLFminJyc/wWIi1NOTo4WLVpU7GMWLVoUtL8kde3a9Zj7S1JBQYF8Pp+qVq1a7P379+/Xrl27gm4AUGo+X+lvAAAAQAnCary3b9+uQ4cOqXbt2kHba9eurfz8/GIfk5+fH9b++/bt03333afevXsf818NRo4cqbS0tMAtMzMznGEAAAAAABAxTn2r+cGDB3X11VfLGKNnn332mPsNGTJEBQUFgdumTZsimCUAAAAAAKFLCGfnmjVrKj4+Xlu2bAnavmXLFqWnpxf7mPT09JD2L2y6N27cqLlz5x73M/JJSUlKSkoKJ3UAiAyuPQcAAMBRwjrjnZiYqJYtW2rOnDmBbX6/X3PmzFH79u2LfUz79u2D9pek2bNnB+1f2HSvXbtWH3zwgWrUqBFOWgAQe7juHAAAIGaEdcZbkgYNGqS+ffuqVatWatOmjcaMGaM9e/aoX79+kqQ+ffqoTp06GjlypCTpjjvuUKdOnfT000+rW7dueu2117R48WI999xzkg433b/73e+0dOlSvfPOOzp06FDg+u/q1asrMTHR1lgBAAAAAIi4sBvva665Rtu2bdOwYcOUn5+v7OxszZw5M/AFanl5eYqL+9+J9A4dOmjy5Ml64IEHNHToUDVu3FhvvfWWmjVrJkn64YcfNH36dElSdnZ20O+aN2+ezj///FIODQAAAACA6Av773i7iL/jLfdzshmLnCIfi5wiH4trxQEAAJzm2d/xBgAAAAAA4aHxBgAAAADAQzTeAAAAAAB4iMYbAAAAAAAP0XgDAAAAAOAhGm8AAAAAADwU9t/xBgCUM/xpMgAAgKjijDcAAAAAAB6i8QYAAAAAwEN81BwAEBo+sg4AAFAqnPEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDCdFOAABwAvL5Sv9YY+zlAQAAEAGc8QYAAAAAwEOc8QYAlG+cPQcAAI7jjDcAAAAAAB6i8QYAAAAAwEM03gAAAAAAeIhrvAEAkLhWHAAAeIYz3gAAAAAAeIjGGwAAAAAAD/FRcwAAbONj6wAA4Aic8QYAAAAAwEOc8QYAwFWcOQcAICbQeAMAcCKgiQcAIGpovAEAQHhsNfH8YwAA4ATBNd4AAAAAAHiIM94AAKD84+w5AMBhnPEGAAAAAMBDnPEGAAAoxJlzAIAHaLwBAAC8QBMPAPj/+Kg5AAAAAAAe4ow3AACA6/gTbgBQrnHGGwAAAAAAD9F4AwAAAADgIT5qDgAAgPDx8XcACBmNNwAAAGIDTTwAR/FRcwAAAAAAPETjDQAAAACAh/ioOQAAAHA0PrYOwCLOeAMAAAAA4CEabwAAAAAAPMRHzQEAAACv8JF1AKLxBgAAAMoH/nY6UG7xUXMAAAAAADzEGW8AAAAApcPZcyAknPEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPFSqxnvixInKyspScnKy2rZtq88///y4+7/xxhs6/fTTlZycrObNm+u9994Lut8Yo2HDhunkk09WxYoVlZOTo7Vr15YmNQAAAAAAnBJ24z1lyhQNGjRIw4cP19KlS9WiRQt17dpVW7duLXb/hQsXqnfv3rrxxhu1bNky9ezZUz179tQ333wT2OeJJ57QuHHjNGnSJH322WeqXLmyunbtqn379pV+ZAAAAAAAOMBnjDHhPKBt27Zq3bq1JkyYIEny+/3KzMzUgAEDNHjw4CL7X3PNNdqzZ4/eeeedwLZ27dopOztbkyZNkjFGGRkZuuuuu3T33XdLkgoKClS7dm299NJLuvbaa4vE3L9/v/bv3x/4uaCgQPXq1dOmTZuUmpoa2kDS0sIZdrCCAvtxbMZyMSebscgp8rHIKfKxyCnyscgp8rHIKfKxyCnyscgp9FhAObNr1y5lZmZq586dSivptWDCsH//fhMfH2+mTZsWtL1Pnz6me/fuxT4mMzPTjB49OmjbsGHDzJlnnmmMMWbdunVGklm2bFnQPh07djS33357sTGHDx9uJHHjxo0bN27cuHHjxo0bN25RvW3atKnEXjpBYdi+fbsOHTqk2rVrB22vXbu2Vq1aVexj8vPzi90/Pz8/cH/htmPtc7QhQ4Zo0KBBgZ/9fr9+/vln1ahRQz6fL5whFVH4rxZhnT33OBY5RT4WOUU+FjlFPhY5RT4WOUU+FjlFPhY5RT4WOUU+FjlFPpbNnGwxxmj37t3KyMgocd+wGm9XJCUlKSkpKWhb1apVrf6O1NRUa0+orVjkFPlY5BT5WOQU+VjkFPlY5BT5WOQU+VjkFPlY5BT5WOQU+Vg2c7KhxI+Y/39hfblazZo1FR8fry1btgRt37Jli9LT04t9THp6+nH3L/xvODEBAAAAACgvwmq8ExMT1bJlS82ZMyewze/3a86cOWrfvn2xj2nfvn3Q/pI0e/bswP4NGjRQenp60D67du3SZ599dsyYAAAAAACUF2F/1HzQoEHq27evWrVqpTZt2mjMmDHas2eP+vXrJ0nq06eP6tSpo5EjR0qS7rjjDnXq1ElPP/20unXrptdee02LFy/Wc889J0ny+XwaOHCgHnnkETVu3FgNGjTQgw8+qIyMDPXs2dPeSEOUlJSk4cOHF/koezRjkVPkY5FT5GORU+RjkVPkY5FT5GORU+RjkVPkY5FT5GORU+Rj2cwpGsL+c2KSNGHCBD355JPKz89Xdna2xo0bp7Zt20qSzj//fGVlZemll14K7P/GG2/ogQce0IYNG9S4cWM98cQTuvTSSwP3G2M0fPhwPffcc9q5c6fOPfdcPfPMMzr11FPLPkIAAAAAAKKoVI03AAAAAAAITVjXeAMAAAAAgPDQeAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADwEI03AAAAAAAeovEGAMSs3Nxc/fbbb1Zi8dc3yy+eOwBAtJ3wf8f7wIEDeuutt7Ro0SLl5+dLktLT09WhQwf16NFDiYmJYcX7/PPPi8Rq37692rRpE3KM7du36x//+EexOd1www2qVatWWDkVMsboww8/1HfffaeTTz5ZXbt2VYUKFUJ+vO1a2cipOA0bNtT777+vxo0bh/yYn376ScuXL1eLFi1UvXp1bd++XS+88IL279+vq666Sk2aNClTToW2bNmiv/3tbxo2bFipY5RmfMYYbdiwQZmZmUpISNCBAwc0bdo07d+/X5deeqlq1qxZ6nxyc3MDz1+zZs2iFuf7779X1apVlZKSErT94MGDWrRokTp27BhSHC/nwoUXXqgXX3xR9evXL3WMzZs3629/+1ugVv3799fpp58e1Zz27Nmj119/PZBT7969VaNGjZAf7+X8TExM1FdffWXlNVyaWCtXrtSnn36q9u3b6/TTT9eqVas0duxY7d+/X3/4wx904YUXhpWDrXk+e/ZsLViwQJ06ddKFF16o+fPna+TIkdq/f7+uv/569evXL6y8ihPueufla6+08+DXX3/VkiVLVL16dTVt2jTovn379un1119Xnz59Qoplcy74/X7FxRU9d+L3+/X999+rXr16UYlVnE2bNmn48OH6xz/+UaY4krRjxw7997//Dbnmkr3XTCGb71vKsnbanJs//vij5syZo+rVqysnJyfo/dyePXv09NNPR/x9y/HYeC9VKJw59e2332rChAnFvs+/7bbbijwPpbVu3Tr96U9/0ty5c8scK5zXn833+TZ7GRu9lStO6Mb7u+++U9euXbV582a1bdtWtWvXlnT4Bf3ZZ5+pbt26mjFjhk455ZQSY23dulVXXnmlPvnkE9WrVy8oVl5ens455xz95z//0UknnXTcOF988YW6du2qSpUqKScnJyjOnDlztHfvXr3//vtq1apViTldeumlevXVV5WWlqaff/5Zl156qT7//HPVrFlTP/30k0499VTNnz8/pMlvq1Y2cxo3blyx2wcNGqR7771X6enpkqTbb7/9uHE+//xzdenSRbt27VLVqlU1e/ZsXXXVVUpISJDf79fmzZu1YMECnX322SXmVJKvvvpKZ599tg4dOlTivrbGt3r1anXt2lWbNm1Sw4YNNWvWLF111VVatWqVjDGqVKmSFi5cGNIB8f/+7//0xBNPKCUlRb/++quuv/56TZs2TcYY+Xw+derUSdOnTy/yBserONLhNww9evTQkiVL5PP5dN111+mZZ54JPHbLli3KyMgIqea25sL06dOL3X7FFVdo7NixyszMlCR17969xJwqVaqkjRs3qlatWvr222/VoUMH1apVS2eddZa+/vpr5eXladGiRTrzzDMjllPTpk21YMECVa9eXZs2bVLHjh21Y8cOnXrqqVq3bp0SEhL06aefqkGDBiXGsjU/r7jiimK3v/3227rwwgtVpUoVSdLUqVNLzGnQoEHFbh87dqz+8Ic/BN4Yjxo16rhxZs6cqR49eiglJUV79+7VtGnT1KdPH7Vo0UJ+v18fffSRZs2aFVLDZXOe/+tf/1K/fv105plnas2aNRo/frzuvPNO/e53v5Pf79e//vUv/fvf/9bvfve7EmMdTzjrna3Xnq3nTpLWrFmjLl26KC8vTz6fT+eee65ee+01nXzyyZLCq7mtubBr1y71799f//3vf5Wamqo///nPGj58uOLj48POyWas4wlnHtiMZes1Y/N9i6210+bc/OKLL9SlSxf5/X4dPHhQderU0VtvvaUzzjgj7Fi23reUJBpzasaMGerZs6fOPvtsde3aNeh98OzZs7VkyRK9/fbb6tq1a8RyshnLZk9kq5ex1Vs5xZzAcnJyTI8ePUxBQUGR+woKCkyPHj1Mly5dQop15ZVXmvbt25tVq1YVuW/VqlWmQ4cO5ne/+12Jcdq2bWtuuukm4/f7i9zn9/vNTTfdZNq1axdSTj6fz2zZssUYY8wtt9ximjZtatavX2+MMWbTpk2mZcuW5uabbw4plq1a2czJ5/OZunXrmqysrKCbz+czderUMVlZWaZBgwYhja1///5m165d5sknnzR169Y1/fv3D9zfr18/07Nnz5By+uqrr457mzJliomLi4vo+Hr06GG6d+9uli9fbgYOHGiaNGlievToYQ4cOGD27dtnLr/8cvOHP/whpJzi4uICz9+QIUNM3bp1zdy5c82ePXvMggULTKNGjczgwYMjFscYY/r06WPatm1rvvjiCzN79mzTsmVL06pVK/Pzzz8bY4zJz883Pp8vpFi25oLP5zNxcXHG5/Md8xbOPCisVY8ePczll19uDh48aIwx5tChQ+baa681l112WdRy+v3vf286dOhgdu7caYwxZvfu3SYnJ8f07t07pFi25qfP5zOdOnUyN9xwQ9AtLi7O9OzZM/BzqOPLzs42559/ftDN5/OZ1q1bm/PPP99ccMEFJcZp3769uf/++40xxrz66qumWrVqZujQoYH7Bw8ebC666KKQcrI5z7Ozs83YsWONMcZ88MEHpmLFimbUqFGB+5966ilzzjnnlBjH5npn87Vn47kzxpiePXuabt26mW3btpm1a9eabt26mQYNGpiNGzcaYw7XPNTx2ZoLt99+uzn11FPNG2+8Yf7+97+b+vXrm27dupn9+/cHcgp1HtiK9fbbbx/3Nnr06JDrVFBQcNzbxx9/HHIsW68Z2+9bbKydNudmTk6O6devnzl06JDZtWuXueWWW0yNGjXM0qVLw45l632LzbXF1pw688wzzYMPPnjM+4cPH26aN28eUk5jx4497u3ee+8NeXy2Xn82eyJbvYyt3solJ3TjXbFiRfP1118f8/7ly5ebihUrhhQrJSUlsEgVZ/HixSYlJaXEOMnJyWblypXHvH/lypUmOTk5pJyOXOBPO+008/bbbwfd/8EHH4S0ABpjr1Y2c/rzn/9ssrOzzbfffhu0PSEhwaxYsSKkGMYYU61atUCMAwcOmLi4OPPZZ58F7l+yZImpU6dOSLGO19wUbg91MbU1vlq1aplly5YZY4z55ZdfjM/nMx9//HHg/k8++cTUq1cvpFhHPn/NmjUzkydPDrr/7bffNqeeemrE4hhjTEZGRtDzVdisZWdnm59++imsNw225sLFF19sunXrFhhjoXCfO2OCa5WZmWnmz58fdP/SpUvNySefHLWcGjZsaGbNmhV0/yeffGIyMzNDimVrfr766qumbt265h//+EfQ9tKMb+TIkaZBgwZmzpw5ZYqVmppq1q5da4w5/I8kCQkJQceJr7/+2tSuXTukWDbneeXKlQONgzHGVKhQwXz11VeBn1euXGlq1KhRYhyb652t156t584YY0466SSzfPnywM9+v9/cfPPNpl69embdunVh1dzWXKhXr56ZN29e4Odt27aZNm3amC5duph9+/aFlZOtWLb/US8uLu6Yt3Bi2XrN2HzfYmvttDk3q1WrZlavXh20beTIkaZatWrm888/DyuWrfctNtcWW3MqOTm52Aaw0KpVq8J6f56RkVHkHygKbxkZGWGPr6yvP5s9ka1exlZv5ZIT+svVqlatqg0bNhzz/g0bNqhq1aohxUpKStKuXbuOef/u3buVlJRUYpz09HR9/vnnx7z/888/D3zUIhQ+n0/S4WtYGjVqFHTfKaecos2bN4cUx2atbOU0adIkDRs2TF27dtWECRNCekxxDhw4oIoVK0qSKlSooEqVKgVdU1r4cbJQVK9eXX//+9+Vm5tb5LZ+/Xq98847Iedla3y//PKLqlevLkmqXLmyKleuHPg4miRlZmZqy5YtIccrfP7y8/OLfLy5RYsW2rRpU0TjFBQUqFq1aoGfk5KSNHXqVGVlZemCCy7Q1q1bQ4oj2ZsLM2bMUOfOndWqVauwnvPi+Hy+QK3i4uKUlpYWdH/VqlW1Y8eOiOZUmJd0+FrCI+eTJNWpU0fbtm0LKY6t+Xnttdfq448/1gsvvKArr7wypJocy+DBgzVlyhTdcsstuvvuu3Xw4MFSxzryuUtOTg56/qpUqaKCgoKQ4tic5xUqVNCBAweCYh15WUdSUpJ+/fXXEuPYXO9svfZsPne//vqrEhISAj/7fD49++yzuvzyy9WpUyetWbMmrHg25sK2bduCvo+hZs2a+uCDD7R7925deuml2rt3b8j52Ip18skna+rUqfL7/cXeli5dGnJOVapU0ciRIzV37txib88991zIsWy+Zmy9bzkyVlnWTttzc9++fUE/Dx48WEOHDlWXLl20cOHCkOPYet9ic22xNaeysrL07rvvHvP+d999N+TvSqlfv75Gjx5d7Phyc3OP+3uOZuv1Z/N9vq1exlZv5ZKEkneJXf3791efPn304IMPqnPnzkWuQXjkkUc0YMCAkGJdc8016tu3r0aPHq3OnTsrNTVV0uFrqObMmaNBgwapd+/eJca5++67ddNNN2nJkiXF5vT3v/9dTz31VMhjvOGGG5SUlKSDBw8qNzc3cM2OdLjhCfVFZLNWtnKSpF69eqlNmzbq06eP3n33Xb344oshP7ZQZmam1q9fr6ysLEkKuk5KOnydWKhf7tSyZUtt3rz5mIvvzp07w/p2XRvjy8jIUF5eXuALcp544omg62G2bdsW9OakJA8++KAqVaqkuLg4bd68Oej5++mnn1S5cuWIxmnYsKGWL18edA1wQkKC3njjDV111VW67LLLQhyZ3blw55136oILLtDvf/97/fe//9Xo0aNDzuNIxhideuqp8vl8+uWXX7R8+fKgf6j47rvvAtfNRSonSercubMSEhK0a9curV69OugL8TZu3BjyFwTZnJ9ZWVmaP3++HnroIbVo0UJ///vfA29yw9W6dWstWbJEt956q1q1aqV///vfYcfKysrS2rVrA2/UFy1aFPRFVXl5eUXeeB+LzXl+yimnaNWqVTrttNMkST/88EPgGnjp8Bf71K1bt8Q4Ntc7m689G8+dJJ1++ulavHhxkS9kK2woQvk+hEK25kK9evW0cuXKoGuAq1SpolmzZqlLly7q1atXyDnZitWyZUstWbJEPXr0KPZ+n88X8jwovIa/U6dOxd5ftWrVkGPZfM3YfN9iY+20OTebNWumhQsXFvkH8Lvvvlt+vz+k965HsvG+xebaYmtOjRgxQtddd50+/PDDYq9bnjlzpiZPnhxSToWvmauvvrrY+8N5zdh6/dl8n2+rl7HVW7nkhG68R4wYocqVK+vJJ5/UXXfdFTgwG2OUnp6u++67T/fee29IsUaNGiW/369rr71Wv/32W+Cb/w4cOKCEhATdeOONIU2yW2+9VTVr1tTo0aP1zDPPBL4MIT4+Xi1bttRLL710zBfq0fr27Rv4/x49ehT51+v//Oc/ys7ODimWrVrZzKlQnTp19MEHH+ixxx7TWWedFfafjbn22muD/uW7W7duQfdPnz495G9OvPnmm7Vnz55j3l+vXr2wD0JlHV9OTo5WrVqlc889V5J0yy23BN0/a9askL84rmPHjlq9erWkw18Ss3HjxqD733vvvaA3JF7HkaRLLrlEzz33nK688sqg7YVvsK688kp9//33IcWyORckKTs7W4sXL9add96p7OzsUv1Jo6Pny9FfbPLpp5+G9WbbRk7Dhw8P+vnoL8H773//q/POOy+kWDbnp3T4bOJDDz2kiy66SH369CnTl9OkpKTon//8p1577TXl5OSEHeuWW24JeszR39Y/Y8aMkL/J2uY8Hzp0aNA/ZhS+mSm0ePHikI4zNtc726+9sj530uEG4tVXX9X1119f5L4JEybI7/dr0qRJIcWyNRe6dOmiF198UZdeemnQ9pSUFL3//vu66KKLQsrHZqx77rnnuPPglFNO0bx580KKdd111x330xbp6elF1p9jsfWasfm+xdbaaXNu9unTRx999JFuvvnmIvfde++9MsaEHKtQWd+32FxbbM2pq666SnXq1NG4ceP09NNPF/mG7Q8//FDt27cPKacRI0Yc9xMlTZs2VW5ubkixbL3+bPZEtnoZW72VS07obzU/Um5ubtCLKJRv4y3Orl27tGTJkqBYLVu2LPLGJhQHDx7U9u3bJR3+CFhxf67i+++/V0ZGRrF/CqQke/bsUXx8vJKTk8N6nK1a2cyp0JIlS7RgwQL16dMnrLO4x7N3717Fx8c78XEWL8aXm5ur5OTkkM+8Hc/69euVmJgY0tkyW3F+++037d2795ivsd9++00//PBDmf5cVqGyzIXp06dr3rx5GjJkiKffwPnJJ5+oVatWIeUYqZzKsk6VZX7+8ssvWrdunZo0aVLkz6CEUyfp8BiWLFminJycIp/GKMv4Qo0TyXl+tHBr5YWyvPYi8dxFKtaOHTuKfELoSLt379bSpUuPeXbPq1jhsFmn44nUa6as71uOJxJrS7jCXQ+8eN9SHthcN23GCmUu2Hyfb6OXsdlbRV1kLykv36pUqWLWrVtnJVazZs1MXl5emePYzMnFWOQU+Vi25qbNWDZzivWax3JOxsT22uliTjZjkVPkY7m4nru4HhgT23PKxZxsxnJxTrlYJ5uxXMzJ5nrglRP6y9XCZSx+OGDDhg1l+tKXQjZzcjEWOUU+lq25aTOWzZxiveaxnJMU22uniznZjEVOkY/l4nru4nogxfaccjEnm7FcnFMu1slmLBdzsrkeeIXGGwAAAAAAD9F4AwAAAADgIRpvAAAAAAA8ROMdhtL+PVgv2czJxVjkFJ1YsSzWa05OoXFxHriYk81Y5BSdWK5xdWyxPKdczMlmLBfnlIt1shnLxZzKAxrvMNj8IgFbXPxyA5uxyCk6sWJZrNecnELj4jxwMSebscgpOrFc4+rYYnlOuZiTzVguzikX62Qzlos5lQul/0L02PH1118f875p06YF/v/jjz82+/btO26sTZs2HfO+RYsWBf7/3//+t/nll1+Oue/cuXOPed+ECRMC/5+Xl2d+++234+Zkc3y2YpFT5GPZmps2Y9nMKdZrPnny5GPed/fddx/3sV7lZHOdspWXrToZY298rq7ntmrl4mvPZs1djOXieu7iemCMm+8RXFxbbK6dLtbcxWOMi8crF+e5zfUg2mi8jTEZGRlm/fr1Rba/+eabplKlSmHFatKkifnpp5+KbF+wYIFJS0sLOU7VqlXN4sWLi2wfM2aMqVKlSlg52RyfrVjkFPlYtuamzVg2c4r1mqelpZn33nuvyPaBAwea9PT0qORkc52ylZetOhljb3yurue2auXia89mzV2M5eJ67uJ6YIyb7xFcXFtsrp0u1tzFY4yLxysX57nN9SDaaLyNMcOGDTMNGzY0P/74Y2Dba6+9ZipVqmRef/31sGL169fPtGzZ0uzatSuw7aOPPjKpqalm1KhRIcf5+9//bmrVqmVWrlwZ2PbUU0+Z1NRUM3/+/LBysjk+W7HIKfKxbM1Nm7Fs5hTrNX/nnXdMWlqa+fjjjwPbbrvtNpORkRG0TkQyJ5vrlK28bNXJGHvjc3U9t1UrF197NmvuYiwX13MX1wNj3HyP4OLaYnPtdLHmLh5jXDxeuTjPba4H0Ubj/f/ddttt5owzzjA//fST+fe//20qVqxo3nzzzbDjHDp0yPTq1ct06tTJ7Nu3z8ydO9ekpKSYMWPGhB3r8ccfN3Xq1DG5ubnmscceM6mpqWbBggVhxzHG3vhsxiKnyMayOTdtxbKZkzGxXXNjDn+Mqlq1ambx4sXmlltuMRkZGWb16tVRzcnWOmUzLxt1KmRrfK6u57Zq5dprzxi7NXctlovruTFurgfGuPkewcW1xeba6VrNXT3GuHi8cm2e214PoonG+wjXXXedady4salUqZJ56623Sh1n//79Jicnx3To0MGkpKSY8ePHlzrWvffea2rUqGGqVq0adB1Dadgan81Y5BTZWDbnpq1YNnMyJrZrbowxEydONElJSaZu3bpm7dq1TuRka52ymZeNOhWyNT5X13NbtXLttWeM3Zq7FsvF9dwYN9cDY9x8j+Di2mJz7XSt5q4eY1w8Xrk2z22vB9HiM+ZE+iq5/5k+fXqRbQcPHtSdd96pLl26qHv37oHtR/5/cZYvX15k2+7du9W7d29169ZNt9xyS2D7mWeeecw448aNK3b7U089pY4dO6pNmzaBbbfffvtxc7I5PluxyCnysWzNTZuxbOYU6zUfNGhQsdvfeOMNnX322WrUqFFg26hRoyKSk811ylZetuok2Rufq+u5rVq5+NqzWXMXY7m4nru4HkhuvkdwcW2xuXa6WHMXjzEuHq9cnOc21wOXnLCNd1xcaH9Jzefz6dChQyXG8vl8QV+Hf+TPhf9fUqwGDRqEnNP69etLzCnUWKGMz0Yscop8LFtz02Ys2zmForzW/IILLgg5p7lz50YkJ9vrlI28bNVJsjc+V9dzm3PKVk62YtmsuYuxXFzPXVwPCmOFmlek3iO4uLbYXDtdrLmLxxgXj1cuznOb64FLTtjG26aNGzeGvG/9+vU9zAQIZnNu2ooV66+XaI/v+++/V0ZGRtCBNNo5HUs08yquTigetXKDi+u5TS7mhKLK03oQK8eY8lRzW2J2PfDqM+yxqFmzZiYvL89KrEsvvdRs3ry5zHGqVKli1q1bZyEju+OzFYucIh/L1ty0GctmTrFec1trgos5GePm2mkrlqvrua28XHztuTgPbMZycT13cT0wxs33CC6uLS6uU7H+/tzFWC7Oc5vrgVdOnH86sWDDhg06ePCglVjz58/Xr7/+WuY4xuIHFmyOz1Yscop8LFtz02YsmznFes1trQku5iS5uXbaiuXqem4rLxdfey7OA5uxXFzPXVwPJDffI7i4tri4TsX6+3MXY7k4z22uB16h8QYAAAAAwEM03gAAAAAAeIjGGwAAAAAAD9F4l3M+ny/aKQBwiItrQqznZCuWi3WS3M3LBhfnge1Yronlsdnm4trC8xcaV2vu4vPnYk5eofEu52x+4QKA8s/FNSHWc3LxC5BscjUvG1ycB7ZjuSaWx2abi2sLz19oXK25i8+fizl5JSHaCZQnf/vb31S7dm0rsYYOHarq1auHtG/hhCzuX4RmzJihOnXqWMnJ5vhsxSKnyMcKZ25GKpbNnGK95t9++60yMjLKHMdmTjbXKVt52aqTFN74yuN6bqtW0Xrt2ax5pJ4/W7GitZ5Hqk4uHhvCjROJWtmsuc2108X3ZS4eY6J1vDoemzV3cb3zis+cSP/McAwjRow47v3Dhg0LOdbLL7983Pv79OkTcqwXXnhBo0eP1tq1ayVJjRs31sCBA9W/f/+QY0h2xydJc+bM0ejRo7Vy5UpJUpMmTTRw4EDl5OSU+NhFixbpp59+0mWXXRbY9vLLL2v48OHas2ePevbsqfHjxyspKSlisWzmdKSy1OlIZX3+1qxZo507d6pNmzZBuT3yyCOB8Q0dOjSsnMo6z73IqTCGjZrbimVzPZCkxYsX6/XXX1deXp4OHDgQdN/UqVOP+1ivav79999r+vTpxeY0atSokOPYrFVZ6nQ0W+NzdT23USubOdmMZavmtmOVZU65uJ4fyVadbOTkxbHd5jHGVq1srVGS3bXTlfeKXrxmbNbJleOVV++FXVvvos6bPw9evmRnZwfdzjjjDFOpUiWTmppqzjrrrLBiVa1aNehWuXJl4/P5TFJSkqlWrVrIcR588EFTuXJlM3jwYPP222+bt99+2wwePNikpKSYBx98MGrjmzhxoklISDDXXnutGTt2rBk7dqzp3bu3qVChgpkwYUKJj7/44ovNY489Fvh5+fLlJiEhwfTv3988/fTTJj093QwfPjykXGzFsplTobLW6Uhlff569uwZNGfWr19vKlasaLp06WJuv/12k5KSYkaPHh1WTmWd517kZLPmtmLZWg+MMebVV181FSpUMJdddplJTEw0l112mTn11FNNWlqaueGGG0p8vBc1/+CDD0ylSpVMs2bNTEJCgsnOzjZVq1Y1aWlp5oILLggrlq1albVOXozP1fXcVq1s5mQrls2a24xV1jnl4npeyGadbORk+9hu8xhjq1Y212Cba6dL7xVtv2Zs1sml45UX74VdXO+ijcb7GAoKCkyvXr3Myy+/XOZYa9asMZ07dzYzZ84M+TE1a9Y0kydPLrJ98uTJpkaNGmXOqbTjq1Onjhk/fnyR7RMmTDAZGRklPj49Pd188cUXgZ+HDh1qzjnnnMDPr7/+umnSpElIudiKZTOnQmWtU0nCef7q1q1rFi5cGPj54YcfNi1atAj8/Pzzzwf9XFrhzHMvcrJZcy+fv9KsB8YY07x588AblpSUFLNu3Trj9/vNn/70JzNs2LASH+9FzVu3bh343YU57d6923Tv3t0888wzYcUqTmlqVdY6HcnW+Fxdz23WylZOtmLZrLnNWGWdUy6u54W8nufh5mT72G7zuGCrVjbXYJvrgUvvFW2/ZmzWyaXjlRfvhcvLehdJNN7HsXz5clO/fn0rsb744gtz2mmnhbx/WlqaWbNmTZHtq1evNmlpaVZyKs34KleubNauXVtk+5o1a0zlypVLfHxSUpLJy8sL/HzOOeeYRx55JPBzbm6uSUlJCSkXW7Fs5lSorHUKRajPX3JyctD4LrzwQvPAAw8Efv7uu++szalQ57kXOdmsudfPX7jrgTHGVKpUyeTm5hpjjKlevbpZvny5McaYb7/91qSnp5f4eC9qnpKSYr777jtjzOEzVN98840xxpgvv/wyamtnWet0JFvjc3U9t1krWznZimWz5jZjlXVOubieF4rEPA8nJ9vHdpvHBVu1srkG21wPXHqvaPs1Y7NOLh2vvHgvXJ7Wu0jhW82Po6CgQAUFBVZiJSQkaPPmzSHvf/311+vZZ58tsv25557T73//eys5lWZ83bt317Rp04psf/vtt4OuCzmW2rVrKzc3V5J04MABLV26VO3atQvcv3v3blWoUCGkXGzFsplTobLWKRShPn/Vq1fXjz/+KEny+/1avHhx0PgOHDhg7RslQ53nXuRks+ZeP3/hrgeSVK1aNe3evVuSVKdOHX3zzTeSpJ07d2rv3r0lPt6LmleuXDlwzdbJJ5+sdevWBe7bvn17WLGOJdxalbVOR7I1PlfXc5u1spWTrVg2a24zVlnnlIvreaFIzPNwcrJ9bLd5XLBVK5trsM31wKX3irZfMzbr5NLxyov3wuVpvYsUvtVc0rhx44J+Nsboxx9/1CuvvKJLLrkkrFjTp08vNtaECRN0zjnnhBXrhRde0KxZswKT7LPPPlNeXp769OmjQYMGBfYr6csJbI6vadOm+utf/6oPP/xQ7du3lyR9+umn+uSTT3TXXXcF/a7bb7+9yOMvvfRSDR48WI8//rjeeustVapUSeedd17g/uXLl6tRo0Yh5WIrls2cCpW1Tkcq6/N3/vnn6+GHH9YzzzyjN954Q36/X+eff37g/m+//VZZWVmhD05ln+de5GSz5rZi2VwPOnbsqNmzZ6t58+a66qqrdMcdd2ju3LmaPXu2OnfuXOLjvah5u3bttGDBAjVp0kSXXnqp7rrrLn399deaOnVq0MExFLZqVdY6Hcnm+Fxcz23VymZONmPZqrnNWGWdUy6u50eyVScbOdk+tts8xkh2amVzjbK5drr0XtH2a8ZmnVw6XnnxXtjF9S7a+FZzSQ0aNAj6OS4uTrVq1dKFF16oIUOGqEqVKiHHiosL/hCBz+cLxHr66ad18sknhxTnggsuCGk/n8+nuXPnHncfm+M7Otbx8lq/fn2R7du3b9cVV1yhBQsWKCUlRf/85z/Vq1evwP2dO3dWu3bt9Ne//rXE32Erls2cCpW1TseLFe7zt2HDBl100UVat26d4uPjNW7cON1yyy2B+3v27KkGDRpo9OjRIeVcmMPR4whnnnuRk5c1L20sW+uBJP3888/at2+fMjIy5Pf79cQTT2jhwoVq3LixHnjgAVWrVu24j/ei5uvXr9cvv/yiM888U3v27NFdd90VyGnUqFGqX79+yLFs1aqsdfJifK6u57Zq5eUxprSxbNbcZqyyzikX1/NCNutkIyfbx3abxxhbtbK5BttcO116r2j7NWOzTi4dr7x4L+ziehdtNN5h+P7775WRkVHkgBBNNnOK5PgKCgqUkpKi+Pj4oO0///yzUlJSlJiYGHJOtmLZzCkajpfXb7/9phUrVqhWrVpF/hbkV199pbp166pGjRqRStXZnGIdNXdbeV3PQ+Xi+FzMqSTl/XUc6blZno/tLuYUSbaeu/L+mokk114vsfbc0XiHITU1VV9++aUaNmzoTCwXc7IZi5wiH4ucIh8rnDhbt27V1q1b5ff7g7afeeaZZcqhLDlJ0i+//FIkp9TUVKs5hZOX7TpFYnzRmpuuzqlIxIpmTrE8p8pbTjZjRSsnW/MpUuuBVL5rbrNOJ8LxysWcooFrvMNg898obMVyMSebscgp8rHIKfKxQomzZMkS9e3bVytXriyyv8/n06FDh6zkEk5Oubm5uu222/Thhx9q3759QY/1IqdQ8rJZp0iPL9Jz08U5FelYkc4p1udUJOO4GiuSOdmcT5FeD6TyWXObdYr145WLOUUbjTcAlAN//OMfdeqpp+qFF15Q7dq15fP5op2S/vCHP8gYo3/84x/O5GSzTi6OzyYX51Ssi/U5hciyOZ9YD0Jjs06xfrxyMadoo/EGgHJg/fr1+s9//qNTTjkl2qkEfPXVV1qyZIlOO+20aKcSYLNOLo7PJhfnVKyL9TmFyLI5n1gPQmOzTrF+vHIxp2g7Mb+tAQDKmc6dO+urr76KdhpBWrdurU2bNkU7jSA26+Ti+GxycU7FulifU4gsm/OJ9SA0NusU68crF3OKNs54h8HmRyRsxXIxJ5uxyCnyscgp8rFCifP888+rb9+++uabb9SsWTNVqFAh6P7u3btbySXcnG6++Wb98MMPxebkxZfxlJSXzTpFenyRnpsuzqlIx4pGzWN5TkUyjquxIpmTzfkU6fVAKr81t3mMieXjlYs5RdsJ23gvX75czZo1C+tr8L3+8gabOTG+yOdUGi4+fyf6l9VEI1YocRYtWqRPPvlEM2bMKHJftL6kZNu2bVq3bp369esXlEs0v1zNZp3KMj4X17ujlaVWLo7PxZyOFunXTHk6rtvKSXLzGOpFTjbnU6SPMVL5mlOFbNYpFo9XtnIqjfLw5WoyJ6i4uDizZcsWY4wxDRo0MNu3by/xMXl5eea3334L+Xf4/X7j9/uLve/jjz82+/bt8ywnxhdaLBfrFIm8SpOT17FczMlmrLLGqV+/vrn11ltNfn5+SL8vEjk1adLEXHHFFebTTz81ubm5ZsOGDUG3aORls05lGZ+L693RylIrF8fnYk5H8+I1E831jmNodHOyOZ+8OMYYEztzqpDNOsXi8cpWTsdic72LhhO28a5evbr59NNPjTHG+Hw+s3XrVmuxn3/+eXPGGWeYxMREk5iYaM444wzz97//PaI5Mb7IxilOaevkZV5lycmrWC7mZDOWrTgpKSnmu+++C/txXuZUqVIls3btWis52crLZp3KMj4X17ujlaVWLo7PxZyOZvM148J6xzE0ujnZnE82105jYm9OFbJZp1g8XtnK6Wg217toOmE/an7llVeqU6dOOvnkk+Xz+dSqVSvFx8cXu+/69etDjjts2DCNGjVKAwYMUPv27SUd/ijJnXfeqby8PI0YMSIiOTG+0GK5WCev8iprTl7EcjEnV8d3xRVXaN68eWrUqFHIj/E6pwsvvFBfffWVlW9ktZWXrTpJZRufi+vd0cpSKxfH52JOR7P1mnFlveMYGt2cbK7BNtfOWJxThWzWKRaPV7ZyOpLN9S7afMaUhw/Ee2PmzJn67rvvdPvtt2vEiBGqUqVKsfvdcccdIcesVauWxo0bp969ewdtf/XVVzVgwABt3749YjkxvtBiuVgnL/KykZPtWC7mZDOWzZz++te/asyYMerWrZuaN29e5EtKbr/99ojn9Nxzz+mRRx7RH//4x2JzCueLYWzlZatOUtnH5+J6d6Sy1srF8bmY05FsvWZcWu84hkYvJ5trsM21M1bnlGS3TrF6vLKVUyGb613URfuUuwtuuOEGs2vXLiux0tLSzJo1a4psX716tUlLS4tKTowvsnGMsVcnm3nZzMlWLBdzshnLZk5ZWVnHvDVo0CAqOfl8vmPe4uLiwoplKy9bdTLG3vhcXO+MsVcrF8fnYk7G2JtTLq53HEMjn5PNNdjm2hnLc8pmnWL9eOXiehdtJ/QZby8MGDBAFSpU0KhRo4K233333fr11181ceLEKGVmR6yPzxYX62QzJ1uxXMzJZqxYnwcnQl6Ai1xc72wip/LNxVq5mBNCE0vP3Ql7jXconnnmGW3fvl3Dhg077n6DBg0K/L/P59Pzzz+vWbNmqV27dpKkzz77THl5eerTp0/EcrIZK9bHZytOJOsUal42c7IVy8WcXB2fLS7m5HJekeLiemeTi+NzMadQubjehYJjqN2cXMWcim3leb1zCWe8j6Nz587Kzc0t8YsELrjggpDi+Xw+zZ07NyI52YwV6+OzFSeSdQo1L5s52YrlYk42Y3k1D448CB0dJzk5Waeccop69Oih6tWrRyyncePGlZhTx44dj/klLV7kVZY6Ha2s4ytJtNc7m7WylZPXsaKdU1nmlIvrXSg4htrN6Ug216iyrgcnypyyuW7G+vHKlfXOJTTeAFAOXHDBBVq6dKkOHTqk0047TZK0Zs0axcfH6/TTT9fq1avl8/m0YMECNW3aNCI5NWjQQNu2bdPevXtVrVo1SdKOHTtUqVIlpaSkaOvWrWrYsKHmzZunzMzMiORks04ujs8mF+dUrIv1OYXIsjmfWA9CY7NOsX68cjGnaIuLdgIAgJL16NFDOTk52rx5s5YsWaIlS5bo+++/10UXXaTevXvrhx9+UMeOHXXnnXdGLKdHH31UrVu31tq1a/XTTz/pp59+0po1a9S2bVuNHTtWeXl5Sk9Pj2hONuvk4vhscnFOxbpYn1OILJvzifUgNDbrFOvHKxdzirpofrObK1566SXzzjvvBH6+5557TFpammnfvr3ZsGGDld8xceJE89BDD0UlJ8YX+ZyOJdw6RSKv0uTkdSwXc7IZqzRxMjIyzIoVK4ps/+abb0xGRoYxxpglS5aYGjVqRCynhg0bmmXLlhXZvnTp0sA3sn7yyScmPT29VDmVJi+bdbI1PhfXO2Ps1crF8bmYkzHev2aiud5xDI18Tjbnk9fHGGNiY07ZrFOsH6/K03oXKTTexphTTz3VzJkzxxhjzMKFC02lSpXM3/72N3P55ZebXr16WfkdF154YVh/GsBmTowv8jkdS7h1ikRepcnJ61gu5mQzVmniVK5c2cybN6/I9nnz5pmUlBRjjDHr1q0zVapUiVhOFStWNF988UWR7Z9//rmpWLGiMcaY3NxcU7ly5VLlVJq8bNbJ1vhcXO+MsVcrF8fnYk7GeP+aieZ6xzE08jnZnE9eH2OMiY05ZbNOsX68Kk/rXaTQeJvDE2Pjxo3GGGPuvfdec/311xtjDv+LU82aNct9Towv8jnZ5GpeiKzrrrvONGjQwEydOtVs2rTJbNq0yUydOtU0bNjQ/OEPfzDGGPPqq6+ali1bRiynSy+91Jx99tlm6dKlgW1Lly41LVu2NN26dTPGGDN9+nTTrFmziOVks062xufiemeMvVq5OD4XczLGzdeMLa4eq1zMy1ZONueTi8cYF587m3WK9eNVLK93pUXjbYypVatWYFJkZ2ebl19+2RhjzHfffVemMzWu5MT4Ip+TTa7mhcjavXu36d+/v0lMTDRxcXEmLi7OJCYmmj/96U/ml19+McYYs2zZsmI/1uWVH3/80eTk5Bifz2cSExMDuV100UUmPz/fGGPM3Llzzfvvvx+xnGzWydb4XFzvjLFXKxfH52JOxrj5mrHF1WOVi3nZysnmfHLxGOPic2ezTrF+vIrl9a60aLzN4X9xOvvss82NN95oKlWqZLZv326MMebtt982Z5xxRlixbF0bYTMnxhf5nGxeI2MrLxeveXQxJ5uxvLg+bffu3earr74yX331ldm9e3fYj/cip5UrV5q3337bvP3222bVqlWlimE7r7LW6UhlHZ+L692RylorF8fnYk5HKuuccnG94xga+ZwK2ViDC9lYO2N9Thlj9xgTq8crWzlF4lr/SKHxNsbs2LHD3HrrraZ79+5mxowZge3Dhg0zjzzySFixbF0bYTMnxhf5nGxeI2MrLxeveXQxJ5uxInF9WrhczMnlvGxwcb2zycXxuZiTTS6udxxDI5+Tq2J9TsUyF+dmLD13/B1vyypVqqRVq1apXr16uu+++/Tjjz/q5Zdf1ooVK3T++edr27Zt0U6xTGJ9fLa4WCebOdmK5WJOLo3viiuu0EsvvaTU1FRdccUVx9136tSpEclp0KBBevjhh1W5cmUNGjTouPuOGjUqpJzKmpfNOnk1Pld4MadwfF7MKRfXO5vI6dhsziev1gNXamUrJ5t1ivXjlevrXbQlRDsBV+zYsUMvvPCCVq5cKUlq0qSJ/vjHP6p69ephxUlJSdFPP/2kevXqadasWYFJl5ycrF9//TUqOdmMFevjc7FOtvKymZOtWC7mZDNWWeOkpaXJ5/MF/t+Gsua0bNkyHTx4MPD/x1KYdyTyslknr8bnynrnxZwqa05exXIlJy/mlIvrncQxNBI52ZxPXq0HsTanbNYp1o9Xrq93URftU+4u+Oijj0xqaqrJzMw0vXr1Mr169TL16tUzqamp5qOPPgorlq1rI2zmxPgin5PNa2Rs5eXiNY8u5mQzlhfXSpWVizm5nJcNLq53Nrk4PhdzssnF9Y5jaORzclWsz6lY5uLcjKXnjsbbGNOsWTPzpz/9yfz222+Bbb/99pu56aabwv6Ke1vXRtjMifFFPieb18jYysvFax5dzMlmLJs57d271+zZsyfw84YNG8zo0aPD/jZQL6/fKigoMNOmTTMrV64M+7G28rJVp+KUdnwurnfG2KuVi+NzMafilHZOubjecQyNfE5HK8sabHPtjOU5ZbNOsX68spWTi9edlxaNtzEmOTm52G/ZW7VqlUlOTo5CRnZzYnyRz8kmV/NCZF100UXm2WefNcYcPgiddNJJpm7duiY5Odk888wzUcnpqquuMuPHjzfGHH4D0bhxY1OhQgWTkJBg3nzzzajkZLNOtsbn4npnjL1auTg+F3Myxs3XjC2uHqtczMtWTjbnk4vHGBefO5t1ivXjVSyvd6XFNd6Szj77bK1cuVKnnXZa0PaVK1eqRYsWYcezcT2KzZwYX+Rzkuxdl2QzL1eueXQ9J5uxbMVZunSpRo8eLUl68803lZ6ermXLluk///mPhg0bpltuuSXiOc2fP1/333+/JGnatGkyxmjnzp365z//qUceeURXXnllWPFs5GWzTrbG5+J6J9mrlYvjczEnye5rxrX1jmNo5HOyOZ9srp1S7M4pm3WK9eOVq+tdVEWz63fFa6+9ZurVq2eefPJJ8/HHH5uPP/7YPPnkkyYrK8u89tprgb+t99VXX5UYy9a1ETZzYnyhxXKxTjbzcvGaRxdzcnV8FStWNBs3bjTGHP5X5L/85S/GGGPy8vJMxYoVo5JTcnKyycvLM8YYc/3115v77rvPGGPMxo0bTeXKlcOKZSsvW3Uyxt74XFzvjLFXKxfH52JOxtibUy6udxxDI5+TzTXY5toZy3PKZp1i/Xjl4noXbTTexhifz3fcW1xcXOC/JbF1bYTNnBhfaLFcrJPNvFy85tHFnGzGsplT8+bNzdixY01eXp5JTU01CxcuNMYYs3jxYlO7du2o5NS4cWMzZcoU88svv5hatWoF/s7ml19+aWrUqBFWLFt52aqTMfbG5+J6Z4y9Wrk4PhdzMsbenHJxveMYGvmcbK7BNtfOWJ5TNusU68crF9e7aOPveEvauHFjyPvWr1//uPdXrFhRX375ZZGPaKxevVrZ2dkhf+29zZwYX2ixXKyTzbxs5mQrlos52YxlM6c333xT1113nQ4dOqTOnTtr1qxZkqSRI0dq/vz5mjFjRsRzeuaZZ3THHXcoJSVF9evX19KlSxUXF6fx48dr6tSpmjdvXsixbOVlq042x+fieifZq5WL43MxJ8nenHJxveMYGvmcbK7BNtfOWJ5TNusU68crF9e7qIt25x9rOnToYKZNm1Zk+7Rp00zbtm0jn5BlsT4+W1ysk82cbMVyMSebsWzPgx9//NEsXbrUHDp0KLDts88+C/qG0E2bNgXd73VOixcvNlOnTjW7d+8ObHvnnXfMggULwopjMy8bdSpka3yuslkrhMbGnHJxvbOJnEJnc42ytR64WCtXjzGxfrxybb2LthP2jPf06dND3rd79+4h7ztlyhTde++9GjBggNq1aydJ+vTTTzVx4kQ99thjatKkSWDfM88807OcGF9osVysk1d5lTUnL2K5mJOr4wtVamqqvvzySzVs2LDc5BSNvELJqSyxXFzvSqu8jM/FnEorkq9jV47rtnLyKi8XcwqV1+vd0WJxToUr0jUvS6wTab2LthO28Y6Liwv62efz6chS+Hy+wP8fOnSo1HGPVvh7fD5fkbg2c2J8ocVysU5e5VXWnLyI5WJONmPZzClUVapU0VdffXXMA5iLOUUjr1ByKkssF9e70iov43Mxp9KK5OvYleO6rZy8ysvFnELl9Xp3tFicU+GKdM3LEutEWu+i7YT9c2J+vz/w/x988IHuu+8+Pfroo2rfvr0kadGiRXrggQf06KOPhhU3NzfXiZwYX2ixXKyTV3mVNScvYrmYk81YNnOyxcWcJHfzKi0X1zubXByfizl5xZX1jmNodHNyVSzOqVjm+tyMqefOu0+xlx9nnHGG+fjjj4tsnz9/vjn99NOjkJHdnBhf5HOyydW84KaUlBSzbt26aKcRJNZzKimWi+tdOMrj+FzMKRwuvmZK4uqxysW8Ip1TJNc7m1x87kLlas0juZ7byimWnLBnvI+0bt06Va1atcj2tLQ0bdiwocTHe3FtRFlzshkr1sdnK45X18iUJS8Xr3l0MSebsaJ9rVRxXMxJcjcv21xc72xycXwu5lRWLq53R+IYGpmcXHUizalY5srcjNnnLtqdvwvOO+88c9FFF5n8/PzAtvz8fNOlSxfTsWPHEh9/rL9xd+TPhbdI5cT4wo/lYp3KmpfNnGzFcjEnV8dXGlWqVCnyL8cu5hTtvI6VkxexXFzvwlEex+diTuHw+nXs4nE91o+htnIqjUisdyfSnApFJI8xNmPF4nrnEhpvY8zatWtNs2bNTGJiomnUqJFp1KiRSUxMNGeccYZZs2ZNWLFmz55tzj77bDNz5kxTUFBgCgoKzMyZM02rVq3MrFmzopIT44t8TrbqZDMvmznZiuViTi6N76uvvgr7zziV9JEtm3WylVNZ8/KiTjZjubTeeVErl8bnck7hiOTr2MXjeqwfQyM9p441n7xaO2NtTtms04l0vLKVUzTet3iFxvv/8/v9ZtasWWbs2LFm7NixZtasWcbv94cdx+a1EbZyshkr1sfnYp1s5eXiNY8u5mQzVlnjxMXFmS1bthhjjGnQoIHZvn17iY/Jy8szv/32m2c5HYvf7z/mnPz444/Nvn37jvv4suTlRZ2OVtbxubLeeVUrV8bnek5HxyztnHJxvTOGY2ikczo6Xrjzyav1INbmlM06nWjHKxs5ledr/Y/GNd6SRowYUWTbokWLtGjRIknSsGHDQo5l69oImzkxvtBiuVgnm3m5eM2jiznZjFXWOFWrVlVubq5OOukkbdiwIeibR48lMzPT05yO9sILL2j06NFau3atJKlx48YaOHCg+vfvH9jn3HPPLTFOWfLyok6FbIzPpfXOi1q5ND6XcypkY065uN5xDI18TlLZ5pNXa2eszSmbdTqRjle2cnLlunMbTti/432ks846K+jngwcPKjc3VwkJCWrUqJGWLl0acqyOHTsqOTlZr7zyimrXri1J2rJli/r06aN9+/bpo48+inhOjC+0WC7WyWZeNnOyFcvFnFwa30033aSXX35ZJ598svLy8lS3bl3Fx8cXu+/69esjOjbp8AF41KhRGjBgQNCfHZkwYYLuvPPOYg/gXuTlRZ1sjs+l9c6LWrk0PpdzkuzNKRfXO46hkc+prPPJq7Uz1uaUzTqdSMcrF9e7qIv2KXdXFRQUmF69epmXX345rMd5eW1EaXOyGSvWx2crjtfXyJQmLxeveXQxJ9fGN2PGDDN+/Hjj8/nMww8/bMaMGVPsLdJjM8aYmjVrmsmTJxfZPnnyZFOjRo2wYpU1L9t1Msbu+I4WzfXOi1qVNadIxHIhJ1tzysX1rjgcQ73NycZ88mI9iMU5ZbNOJ8rxysX1Lto4430cX3/9tS6//PKwP8ZgjNEHH3yglStXSpKaNGminJwc+Xy+qOVkM1asj89WHC/rVNq8bOZkK5aLOdmMZStOv379NG7cOFWpUiWsx3mZU9WqVfXFF1+ocePGQdvXrFmjNm3aaOfOnRHPy2adbI/vaNFe72zWylZOXseKdk4255SL611xOIZ6l5PN+WR7PYjVOWWzTrF+vHJ1vYsmGu/jWLBggS6//HLt2LEj5MeU9LGJcK+NsJGTzVixPj5bcbyukxR+XjZzshXLxZxsxorEPAiXzZwGDBigChUqaNSoUUHb7777bv3666+aOHFiVPKyxeb4iuPiemeTi+OLdk625pSL692xcAz1Liev16jSOlHnVDS5eLxycb2LNhpvSePGjQv62RijH3/8Ua+88oo6deqkyZMnhxzL1rURNnNifKHFcrFONvNy8ZpHF3OyGctmTsfyzDPPaPv27SEfeMqa06BBgwL//9tvv+mll15SvXr11K5dO0nSZ599pry8PPXp00fjx48PeRxe1yrUOnkxPhfXu+MJd065OD6XcvJiTrm43nEMjUxOXq3BxxLueiDF/pwqTmnqVNZYLh6vXF/voo3GW1KDBg2Cfo6Li1OtWrV04YUXasiQIWX+CMiuXbt0ww03qFevXrr++usjnhPjCy2Wi3XyOq/S5uRlLBdzshnLZk6S1LlzZ+Xm5ob1JSxlyemCCy4IKabP59PcuXNLnVO4eZUk1Dp5MT4X17vjCXdOuTg+l3KK1Gsm2usdx9DI5BTJNViyc4yRYmtOFcdWncKJ5eLxqjyudxHlzaXjONry5ctN/fr1o52GZ2J9fLa4WCebOdmK5WJONmPF+jywydW8ABe5uN7ZRE7lm4u1cjEnhKY8Pndx0W78TxQFBQUqKCiIdhqeifXx2eJinWzmZCuWiznZjBXr88AmV/MCXOTiemcTOZVvLtbKxZwQmvL43CVEO4FYc7xrIy655JIoZWVPrI/PFhfrZDMnW7FczMlmLJs5/fOf/1TNmjXVrVs3SdK9996r5557Tk2bNtWrr76q+vXrRzynYynNtW628rJVp+OxeS1fNEWiVgiNzevqo7l22kJOZVOaNcrmeuBirVw8xpyox6tornfRxjXelkXi+rtoivXx2eJinVy65tHlnGzGspnTaaedpmeffVYXXnihFi1apJycHI0ePVrvvPOOEhISNHXq1IjndCyludbNVl626nQ8Nq/li6ZI1AqhiYXr6m0ip7IpzRplcz1wsVYuHmNO1ONVNNe7aKPxBoByoFKlSlq1apXq1aun++67Tz/++KNefvllrVixQueff762bdsW7RSdQJ1CR60AFGI9CI3NOlHzEw/XeANAOZCSkqKffvpJkjRr1ixddNFFkqTk5GT9+uuv0UzNKdQpdNQKQCHWg9DYrBM1P/HQeANAOXDRRRepf//+6t+/v9asWaNLL71UkrRixQplZWVFJad//vOfevfddwM/33vvvapatao6dOigjRs3RiUnm3VycXw2uTinYl2szylEls35xHoQGpt1ivXjlYs5RRuNNwCUAxMnTlT79u21bds2/ec//1GNGjUkSUuWLFHv3r2jktOjjz6qihUrSpIWLVqkiRMn6oknnlDNmjV15513RiUnm3VycXw2uTinYl2szylEls35xHoQGpt1ivXjlYs5RRvXeAMASiXWr0+L9fEh8phTsIn5hEIuzgUXc4o2/pwYAJQTO3bs0AsvvKCVK1dKkpo0aaI//vGPql69elTyKbw+rV69epo1a5YGDRokKfrXp9mqk6vjs8m1ORXrToQ5hcixPZ9YD0Jjs06xfLxyMado46PmAFAOzJ8/X1lZWRo3bpx27NihHTt2aPz48WrQoIHmz58flZxcvCbQZp1cHJ9NLs6pWBfrcwqRZXM+sR6ExmadYv145WJO0cZHzQGgHGjevLnat2+vZ599VvHx8ZKkQ4cO6f/+7/+0cOFCff311xHPaefOnXrggQe0adMm3XLLLbr44oslScOHD1diYqLuv//+iOdks04ujs8mF+dUrIv1OYXIsjmfWA9CY7NOsX68cjGnaKPxBoByoGLFivryyy912mmnBW1fvXq1srOzT9iPbR2NOoWOWgEoxHoQGpt1ouYnHq7xBoBy4Oyzz9bKlSuLHKBXrlypFi1aRCkr964JtF0n18Znk6tzKtbF8pxC5NmaT6wHobFZpxPheOViTtHEGW8AKAemTJmie++9VwMGDFC7du0kSZ9++qkmTpyoxx57TE2aNAnse+aZZ0Ykp/nz5+vyyy9XWlqaWrVqJenwn0HZuXOn/vvf/6pjx44RyeNINuvk4vhscnFOxbpYn1OILJvzifUgNDbrFOvHKxdzijYabwAoB+Lijv9dmD6fT8YY+Xw+HTp0KCI5uXhNoM06uTg+m1ycU7Eu1ucUIsvmfGI9CI3NOsX68crFnKKNxhsAyoGNGzeGvG/9+vU9zOR/XLw+zWadXByfTS7OqVgX63MKkWVzPrEehMZmnWL9eOViTtHGNd4AUA64+EbHxWsCbdbJxfHZ5OKcinWxPqcQWTbnE+tBaGzWKdaPVy7mFG003gDgqOnTp4e8b/fu3T3MpHi333677rjjDn333XfFXp+2fPnywL5eXhPoVZ1cGZ9Nrs+pWBeLcwrRU9b5xHoQGpt1OpGOVy7mFG181BwAHHX09V+F13sd+XOhaFxz58o1gV7VyZXx2eT6nIp1sTinED1lnU+sB6GxWacT6XjlYk7RxhlvAHCU3+8P/P8HH3yg++67T48++qjat28vSVq0aJEeeOABPfroo1HJLzc3Nyq/92he1cmV8dnk+pyKdbE4pxA9ZZ1PrAehsVmnE+l45WJO0cYZbwAoB5o1a6ZJkybp3HPPDdr+8ccf66abbgr8jcwTHXUKHbUCUIj1IDQ260TNTzyc8QaAcmDdunWqWrVqke1paWnasGFDxPJw/ZrAstbJ9fHZ5MqcinUn0pyC97yaT6wHobFZp1g8XrmYk0s44w0A5UDHjh2VnJysV155RbVr15YkbdmyRX369NG+ffv00UcfRSQP168JLGudXB+fTa7MqVh3Is0peM+r+cR6EBqbdYrF45WLObnk+Fe9AwCc8I9//EM//vij6tWrp1NOOUWnnHKK6tWrpx9++EHPP/98xPLw+/2B26xZs5Sdna0ZM2Zo586d2rlzp9577z2dffbZmjlzZsRyOlJZ6+T6+GxyZU7FuhNpTsF7Xs0n1oPQ2KxTLB6vXMzJJZzxBoBywhijDz74IHDdV5MmTZSTkxP0L8iR5Or1abbq5Or4bHJtTsW6E2FOIXJszyfWg9DYrFMsH69czCnauMYbAMqBESNGFNm2aNEiLVq0SJI0bNiwSKfk5DWBNuvk4vhscnFOxbpYn1OILJvzifUgNDbrFOvHKxdzijbOeANAOXDWWWcF/Xzw4EHl5uYqISFBjRo10tKlSyOek4vXBNqsk4vjs8nFORXrYn1OIbJszifWg9DYrFOsH69czCnaOOMNAOXAsmXLimzbtWuXbrjhBvXq1SsKGR2+Pq1Xr16qV6+eMjMzJUmbNm1S48aNNW3atKjkZLNOLo7PJhfnVKyL9TmFyLI5n1gPQmOzTrF+vHIxp2jjjDcAlGNff/21Lr/88qh9bKu8XBNY2jqVl/HZFO05FetOxDkF73g9n1gPQmOzTrF0vHIxp2jijDcAlGMFBQUqKCiIyu8uT9cElqZO5Wl8NkVzTsW6E3VOwRuRmE+sB6GxWadYOV65mFO0ccYbAMqBcePGBf1sjNGPP/6oV155RZ06ddLkyZMjnpOL1wTarJOL47PJxTkV62J9TiGybM4n1oPQ2KxTrB+vXMwp2mi8AaAcaNCgQdDPcXFxqlWrli688EINGTJEVapUiVJmwY68Pu3666+P+O/3uk7RHp9N5WVOxbpYmlOIvtLOJ9aD0Nis04l4vHIxp0ii8QYAWBXr1wTG+vgQecwp2MR8QiEX54KLOUVKXLQTAADElli/JjDWx4fIY07BJuYTCrk4F1zMKVL4cjUAQKkc7/q0Sy65JEpZ2RPr40PkMadgE/MJhVycCy7mFG181BwAUCqxfk1grI8Pkcecgk3MJxRycS64mFO00XgDAAAAAOAhrvEGAAAAAMBDNN4AAAAAAHiIxhsAAAAAAA/ReAMAAAAA4CEabwAAAAAAPETjDQAAAACAh2i8AQAAAADw0P8DzdI+eHJuW3EAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 1200x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#  feature importances\n",
    "importances = clf.feature_importances_\n",
    "indices = np.argsort(importances)[::-1]\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "plt.title(\"Feature importances\")\n",
    "plt.bar(range(X_train.shape[1]), importances[indices],\n",
    "        color=\"r\", align=\"center\")\n",
    "plt.xticks(range(X_train.shape[1]), X_train.columns[indices], rotation=90)\n",
    "plt.xlim([-1, X_train.shape[1]+1])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1d89b9de-19ce-4293-b3ae-afabe3a67cb4",
   "metadata": {},
   "source": [
    "### SVM"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e097c028-e83f-4496-83a8-d607bc93296e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Train Accuracy: 0.88014052987463\n",
      "Test Accuracy: 0.8772858517805582\n",
      "\n",
      "Confusion Matrix:\n",
      "[[0.92156863 0.07843137]\n",
      " [0.35410425 0.64589575]]\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "       NoVPN       0.93      0.92      0.93      8721\n",
      "         VPN       0.61      0.65      0.63      1669\n",
      "\n",
      "    accuracy                           0.88     10390\n",
      "   macro avg       0.77      0.78      0.78     10390\n",
      "weighted avg       0.88      0.88      0.88     10390\n",
      "\n",
      "\n",
      "Detailed Misclassification Summary (per Category):\n",
      "            category true_label predicted_label  count\n",
      "0               Chat      NoVPN             VPN    574\n",
      "1               Chat        VPN           NoVPN    401\n",
      "2  Command & Control      NoVPN             VPN     19\n",
      "3  Command & Control        VPN           NoVPN     97\n",
      "4      File Transfer      NoVPN             VPN     13\n",
      "5      File Transfer        VPN           NoVPN     43\n",
      "6          Streaming      NoVPN             VPN     35\n",
      "7          Streaming        VPN           NoVPN     23\n",
      "8               VoIP      NoVPN             VPN     43\n",
      "9               VoIP        VPN           NoVPN     27\n",
      "Confusion matrix saved at: output_results\\confusion_matrix_SVM5.png\n",
      "Accuracy and classification report saved at: output_results\\SVM5.txt\n",
      "Detailed misclassification summary saved at: output_results\\SVM5.csv\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.svm import LinearSVC\n",
    "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "\n",
    "output_dir = \"output_results\"\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "\n",
    "file_path = './VNAT/WF/flows_extra_labelled_WF5.csv'  \n",
    "df = pd.read_csv(file_path)\n",
    "\n",
    "df['label_encoded'] = df['label'].map({'NoVPN': 0, 'VPN': 1})\n",
    "\n",
    "category = df['category']\n",
    "\n",
    "X = df.drop(['label', 'label_encoded', 'category'], axis=1)\n",
    "y = df['label_encoded']\n",
    "\n",
    "X_train, X_test, y_train, y_test, category_train, category_test = train_test_split(\n",
    "    X, y, category, test_size=0.2, random_state=42\n",
    ")\n",
    "\n",
    "scaler = StandardScaler()\n",
    "X_train = pd.DataFrame(scaler.fit_transform(X_train), columns=X.columns, index=X_train.index)\n",
    "X_test = pd.DataFrame(scaler.transform(X_test), columns=X.columns, index=X_test.index)\n",
    "\n",
    "clf = LinearSVC(random_state=42, max_iter=10000)\n",
    "clf.fit(X_train, y_train)\n",
    "\n",
    "y_pred = clf.predict(X_test)\n",
    "\n",
    "train_accuracy = accuracy_score(y_train, clf.predict(X_train))\n",
    "test_accuracy = accuracy_score(y_test, y_pred)\n",
    "\n",
    "cm = confusion_matrix(y_test, y_pred, normalize='true')\n",
    "\n",
    "print(\"Train Accuracy:\", train_accuracy)\n",
    "print(\"Test Accuracy:\", test_accuracy)\n",
    "print(\"\\nConfusion Matrix:\")\n",
    "print(cm)\n",
    "print(\"\\nClassification Report:\")\n",
    "print(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "plt.figure(figsize=(6, 5))\n",
    "sns.heatmap(cm, annot=True, fmt=\".2f\", cmap=\"Blues\", xticklabels=['NoVPN', 'VPN'], \n",
    "yticklabels=['NoVPN', 'VPN'], annot_kws={\"size\": 16})  # Enlarged numbers inside the matrix\n",
    "\n",
    "plt.xlabel(\"Predicted Label\", fontsize=20)\n",
    "plt.ylabel(\"True Label\", fontsize=20)\n",
    "plt.title(\"\", fontsize=20)\n",
    "plt.xticks(fontsize=20)\n",
    "plt.yticks(fontsize=20)\n",
    "plt.tight_layout()\n",
    "\n",
    "conf_matrix_path = os.path.join(output_dir, \"confusion_matrix_SVM5.png\")\n",
    "plt.savefig(conf_matrix_path, dpi=300, bbox_inches='tight')\n",
    "plt.close()\n",
    "\n",
    "accuracy_report_path = os.path.join(output_dir, \"SVM5.txt\")\n",
    "with open(accuracy_report_path, \"w\") as f:\n",
    "    f.write(f\"Train Accuracy: {train_accuracy:.4f}\\n\")\n",
    "    f.write(f\"Test Accuracy: {test_accuracy:.4f}\\n\\n\")\n",
    "    f.write(\"Classification Report:\\n\")\n",
    "    f.write(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "misclassified_mask = (y_test != y_pred)\n",
    "misclassified_samples = X_test.loc[misclassified_mask].copy()\n",
    "\n",
    "misclassified_samples['true_label'] = y_test.loc[misclassified_mask]\n",
    "misclassified_samples['predicted_label'] = y_pred[misclassified_mask]\n",
    "misclassified_samples['category'] = category_test.loc[misclassified_mask]\n",
    "\n",
    "label_map = {0: 'NoVPN', 1: 'VPN'}\n",
    "misclassified_samples['true_label'] = misclassified_samples['true_label'].map(label_map)\n",
    "misclassified_samples['predicted_label'] = misclassified_samples['predicted_label'].map(label_map)\n",
    "\n",
    "misclassification_detail = misclassified_samples.groupby(['category', 'true_label', 'predicted_label']).size().reset_index(name='count')\n",
    "\n",
    "print(\"\\nDetailed Misclassification Summary (per Category):\")\n",
    "print(misclassification_detail)\n",
    "\n",
    "misclassification_detail_path = os.path.join(output_dir, \"SVM5.csv\")\n",
    "misclassification_detail.to_csv(misclassification_detail_path, index=False)\n",
    "\n",
    "print(\"Confusion matrix saved at:\", conf_matrix_path)\n",
    "print(\"Accuracy and classification report saved at:\", accuracy_report_path)\n",
    "print(\"Detailed misclassification summary saved at:\", misclassification_detail_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "bbd06971-d739-43ef-bca3-3599021f5404",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "C:\\Users\\yasah\\anaconda3\\envs\\py311\\Lib\\site-packages\\keras\\src\\layers\\core\\dense.py:87: UserWarning: Do not pass an `input_shape`/`input_dim` argument to a layer. When using Sequential models, prefer using an `Input(shape)` object as the first layer in the model instead.\n",
      "  super().__init__(activity_regularizer=activity_regularizer, **kwargs)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m6s\u001b[0m 3ms/step - accuracy: 0.8774 - loss: 0.2328 - val_accuracy: 0.8874 - val_loss: 0.1910\n",
      "Epoch 2/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.8920 - loss: 0.1884 - val_accuracy: 0.9080 - val_loss: 0.1785\n",
      "Epoch 3/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9067 - loss: 0.1755 - val_accuracy: 0.9126 - val_loss: 0.1696\n",
      "Epoch 4/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m5s\u001b[0m 3ms/step - accuracy: 0.9154 - loss: 0.1660 - val_accuracy: 0.9190 - val_loss: 0.1639\n",
      "Epoch 5/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.9207 - loss: 0.1598 - val_accuracy: 0.9171 - val_loss: 0.1641\n",
      "Epoch 6/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.9206 - loss: 0.1582 - val_accuracy: 0.9195 - val_loss: 0.1600\n",
      "Epoch 7/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m4s\u001b[0m 3ms/step - accuracy: 0.9288 - loss: 0.1477 - val_accuracy: 0.9247 - val_loss: 0.1558\n",
      "Epoch 8/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9297 - loss: 0.1466 - val_accuracy: 0.9248 - val_loss: 0.1541\n",
      "Epoch 9/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 3ms/step - accuracy: 0.9312 - loss: 0.1458 - val_accuracy: 0.9291 - val_loss: 0.1494\n",
      "Epoch 10/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9337 - loss: 0.1414 - val_accuracy: 0.9302 - val_loss: 0.1502\n",
      "Epoch 11/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9356 - loss: 0.1385 - val_accuracy: 0.9297 - val_loss: 0.1487\n",
      "Epoch 12/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9345 - loss: 0.1352 - val_accuracy: 0.9280 - val_loss: 0.1522\n",
      "Epoch 13/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9382 - loss: 0.1335 - val_accuracy: 0.9294 - val_loss: 0.1522\n",
      "Epoch 14/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9362 - loss: 0.1341 - val_accuracy: 0.9295 - val_loss: 0.1504\n",
      "Epoch 15/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9372 - loss: 0.1300 - val_accuracy: 0.9291 - val_loss: 0.1482\n",
      "Epoch 16/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9357 - loss: 0.1343 - val_accuracy: 0.9324 - val_loss: 0.1442\n",
      "Epoch 17/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m2s\u001b[0m 2ms/step - accuracy: 0.9406 - loss: 0.1298 - val_accuracy: 0.9332 - val_loss: 0.1430\n",
      "Epoch 18/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9436 - loss: 0.1227 - val_accuracy: 0.9297 - val_loss: 0.1449\n",
      "Epoch 19/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9397 - loss: 0.1295 - val_accuracy: 0.9311 - val_loss: 0.1460\n",
      "Epoch 20/20\n",
      "\u001b[1m1299/1299\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m3s\u001b[0m 2ms/step - accuracy: 0.9426 - loss: 0.1237 - val_accuracy: 0.9340 - val_loss: 0.1432\n",
      "\u001b[1m325/325\u001b[0m \u001b[32m━━━━━━━━━━━━━━━━━━━━\u001b[0m\u001b[37m\u001b[0m \u001b[1m0s\u001b[0m 988us/step\n",
      "Train Accuracy: 0.9446302652359009\n",
      "Test Accuracy: 0.9339749759384023\n",
      "\n",
      "Confusion Matrix:\n",
      "[[0.95986699 0.04013301]\n",
      " [0.20131815 0.79868185]]\n",
      "\n",
      "Classification Report:\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "       NoVPN       0.96      0.96      0.96      8721\n",
      "         VPN       0.79      0.80      0.80      1669\n",
      "\n",
      "    accuracy                           0.93     10390\n",
      "   macro avg       0.88      0.88      0.88     10390\n",
      "weighted avg       0.93      0.93      0.93     10390\n",
      "\n",
      "\n",
      "Detailed Misclassification Summary (per Category):\n",
      "            category true_label predicted_label  count\n",
      "0               Chat      NoVPN             VPN    297\n",
      "1               Chat        VPN           NoVPN    277\n",
      "2  Command & Control      NoVPN             VPN     30\n",
      "3  Command & Control        VPN           NoVPN     27\n",
      "4      File Transfer      NoVPN             VPN      2\n",
      "5      File Transfer        VPN           NoVPN     19\n",
      "6          Streaming      NoVPN             VPN     18\n",
      "7          Streaming        VPN           NoVPN      6\n",
      "8               VoIP      NoVPN             VPN      3\n",
      "9               VoIP        VPN           NoVPN      7\n",
      "Confusion matrix saved at: output_results\\confusion_matrix_NN5.png\n",
      "Accuracy and classification report saved at: output_results\\NN5.txt\n",
      "Detailed misclassification summary saved at: output_results\\NN5.csv\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import StandardScaler\n",
    "from sklearn.metrics import classification_report, confusion_matrix, accuracy_score\n",
    "import tensorflow as tf\n",
    "from tensorflow import keras\n",
    "\n",
    "output_dir = \"output_results\"\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "\n",
    "file_path = './VNAT/WF/flows_extra_labelled_WF5.csv'  \n",
    "df = pd.read_csv(file_path)\n",
    "\n",
    "df['label_encoded'] = df['label'].map({'NoVPN': 0, 'VPN': 1})\n",
    "\n",
    "category = df['category']\n",
    "\n",
    "X = df.drop(['label', 'label_encoded', 'category'], axis=1)\n",
    "y = df['label_encoded']\n",
    "\n",
    "X_train, X_test, y_train, y_test, category_train, category_test = train_test_split(\n",
    "    X, y, category, test_size=0.2, random_state=42\n",
    ")\n",
    "\n",
    "scaler = StandardScaler()\n",
    "X_train = scaler.fit_transform(X_train)\n",
    "X_test = scaler.transform(X_test)\n",
    "\n",
    "model = keras.Sequential([\n",
    "    keras.layers.Dense(64, activation='relu', input_shape=(X_train.shape[1],)),\n",
    "    keras.layers.Dense(32, activation='relu'),\n",
    "    keras.layers.Dense(1, activation='sigmoid')\n",
    "])\n",
    "\n",
    "model.compile(optimizer='adam', loss='binary_crossentropy', metrics=['accuracy'])\n",
    "\n",
    "model.fit(X_train, y_train, epochs=20, batch_size=32, validation_data=(X_test, y_test), verbose=1)\n",
    "\n",
    "y_pred_probs = model.predict(X_test)\n",
    "y_pred = (y_pred_probs > 0.5).astype(int).flatten()\n",
    "\n",
    "train_accuracy = model.evaluate(X_train, y_train, verbose=0)[1]\n",
    "test_accuracy = accuracy_score(y_test, y_pred)\n",
    "\n",
    "cm = confusion_matrix(y_test, y_pred, normalize='true')\n",
    "\n",
    "print(\"Train Accuracy:\", train_accuracy)\n",
    "print(\"Test Accuracy:\", test_accuracy)\n",
    "print(\"\\nConfusion Matrix:\")\n",
    "print(cm)\n",
    "print(\"\\nClassification Report:\")\n",
    "print(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "plt.figure(figsize=(6, 5))\n",
    "sns.heatmap(cm, annot=True, fmt=\".2f\", cmap=\"Blues\", xticklabels=['NoVPN', 'VPN'], \n",
    "yticklabels=['NoVPN', 'VPN'], annot_kws={\"size\": 16})  # Enlarged numbers inside the matrix\n",
    "\n",
    "plt.xlabel(\"Predicted Label\", fontsize=20)\n",
    "plt.ylabel(\"True Label\", fontsize=20)\n",
    "plt.title(\"\", fontsize=20)\n",
    "plt.xticks(fontsize=20)\n",
    "plt.yticks(fontsize=20)\n",
    "plt.tight_layout()\n",
    "\n",
    "conf_matrix_path = os.path.join(output_dir, \"confusion_matrix_NN5.png\")\n",
    "plt.savefig(conf_matrix_path, dpi=300, bbox_inches='tight')\n",
    "plt.close()\n",
    "\n",
    "accuracy_report_path = os.path.join(output_dir, \"NN5.txt\")\n",
    "with open(accuracy_report_path, \"w\") as f:\n",
    "    f.write(f\"Train Accuracy: {train_accuracy:.4f}\\n\")\n",
    "    f.write(f\"Test Accuracy: {test_accuracy:.4f}\\n\\n\")\n",
    "    f.write(\"Classification Report:\\n\")\n",
    "    f.write(classification_report(y_test, y_pred, target_names=['NoVPN', 'VPN']))\n",
    "\n",
    "misclassified_mask = (y_test != y_pred)\n",
    "misclassified_samples = X_test[misclassified_mask]\n",
    "\n",
    "misclassified_samples = pd.DataFrame(misclassified_samples, columns=X.columns)\n",
    "misclassified_samples['true_label'] = y_test.loc[misclassified_mask].values\n",
    "misclassified_samples['predicted_label'] = y_pred[misclassified_mask]\n",
    "misclassified_samples['category'] = category_test.loc[misclassified_mask].values\n",
    "\n",
    "label_map = {0: 'NoVPN', 1: 'VPN'}\n",
    "misclassified_samples['true_label'] = misclassified_samples['true_label'].map(label_map)\n",
    "misclassified_samples['predicted_label'] = misclassified_samples['predicted_label'].map(label_map)\n",
    "\n",
    "misclassification_detail = misclassified_samples.groupby(['category', 'true_label', 'predicted_label']).size().reset_index(name='count')\n",
    "\n",
    "print(\"\\nDetailed Misclassification Summary (per Category):\")\n",
    "print(misclassification_detail)\n",
    "\n",
    "misclassification_detail_path = os.path.join(output_dir, \"NN5.csv\")\n",
    "misclassification_detail.to_csv(misclassification_detail_path, index=False)\n",
    "\n",
    "print(\"Confusion matrix saved at:\", conf_matrix_path)\n",
    "print(\"Accuracy and classification report saved at:\", accuracy_report_path)\n",
    "print(\"Detailed misclassification summary saved at:\", misclassification_detail_path)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "159bf0ae-4567-4f51-ab97-df217f5e0dc9",
   "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": "3.11.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}