DL_A4_43217.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "djQINsl_Ab3V"
   },
   "source": [
    "Name: **Megha Dandapat**<br>\n",
    "Div: **BE10**<br>\n",
    "Roll no: **43217**<br>\n",
    "Title: **Assignment 4: ECG Anomaly detection using Autoencoders**<br>"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 299
    },
    "id": "HnMj8dfZDIWx",
    "outputId": "94e6d5c9-ab55-40b9-a9ad-dcbaf9ac5191"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "      <th>4</th>\n",
       "      <th>5</th>\n",
       "      <th>6</th>\n",
       "      <th>7</th>\n",
       "      <th>8</th>\n",
       "      <th>9</th>\n",
       "      <th>...</th>\n",
       "      <th>131</th>\n",
       "      <th>132</th>\n",
       "      <th>133</th>\n",
       "      <th>134</th>\n",
       "      <th>135</th>\n",
       "      <th>136</th>\n",
       "      <th>137</th>\n",
       "      <th>138</th>\n",
       "      <th>139</th>\n",
       "      <th>140</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-0.112522</td>\n",
       "      <td>-2.827204</td>\n",
       "      <td>-3.773897</td>\n",
       "      <td>-4.349751</td>\n",
       "      <td>-4.376041</td>\n",
       "      <td>-3.474986</td>\n",
       "      <td>-2.181408</td>\n",
       "      <td>-1.818286</td>\n",
       "      <td>-1.250522</td>\n",
       "      <td>-0.477492</td>\n",
       "      <td>...</td>\n",
       "      <td>0.792168</td>\n",
       "      <td>0.933541</td>\n",
       "      <td>0.796958</td>\n",
       "      <td>0.578621</td>\n",
       "      <td>0.257740</td>\n",
       "      <td>0.228077</td>\n",
       "      <td>0.123431</td>\n",
       "      <td>0.925286</td>\n",
       "      <td>0.193137</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-1.100878</td>\n",
       "      <td>-3.996840</td>\n",
       "      <td>-4.285843</td>\n",
       "      <td>-4.506579</td>\n",
       "      <td>-4.022377</td>\n",
       "      <td>-3.234368</td>\n",
       "      <td>-1.566126</td>\n",
       "      <td>-0.992258</td>\n",
       "      <td>-0.754680</td>\n",
       "      <td>0.042321</td>\n",
       "      <td>...</td>\n",
       "      <td>0.538356</td>\n",
       "      <td>0.656881</td>\n",
       "      <td>0.787490</td>\n",
       "      <td>0.724046</td>\n",
       "      <td>0.555784</td>\n",
       "      <td>0.476333</td>\n",
       "      <td>0.773820</td>\n",
       "      <td>1.119621</td>\n",
       "      <td>-1.436250</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-0.567088</td>\n",
       "      <td>-2.593450</td>\n",
       "      <td>-3.874230</td>\n",
       "      <td>-4.584095</td>\n",
       "      <td>-4.187449</td>\n",
       "      <td>-3.151462</td>\n",
       "      <td>-1.742940</td>\n",
       "      <td>-1.490659</td>\n",
       "      <td>-1.183580</td>\n",
       "      <td>-0.394229</td>\n",
       "      <td>...</td>\n",
       "      <td>0.886073</td>\n",
       "      <td>0.531452</td>\n",
       "      <td>0.311377</td>\n",
       "      <td>-0.021919</td>\n",
       "      <td>-0.713683</td>\n",
       "      <td>-0.532197</td>\n",
       "      <td>0.321097</td>\n",
       "      <td>0.904227</td>\n",
       "      <td>-0.421797</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.490473</td>\n",
       "      <td>-1.914407</td>\n",
       "      <td>-3.616364</td>\n",
       "      <td>-4.318823</td>\n",
       "      <td>-4.268016</td>\n",
       "      <td>-3.881110</td>\n",
       "      <td>-2.993280</td>\n",
       "      <td>-1.671131</td>\n",
       "      <td>-1.333884</td>\n",
       "      <td>-0.965629</td>\n",
       "      <td>...</td>\n",
       "      <td>0.350816</td>\n",
       "      <td>0.499111</td>\n",
       "      <td>0.600345</td>\n",
       "      <td>0.842069</td>\n",
       "      <td>0.952074</td>\n",
       "      <td>0.990133</td>\n",
       "      <td>1.086798</td>\n",
       "      <td>1.403011</td>\n",
       "      <td>-0.383564</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.800232</td>\n",
       "      <td>-0.874252</td>\n",
       "      <td>-2.384761</td>\n",
       "      <td>-3.973292</td>\n",
       "      <td>-4.338224</td>\n",
       "      <td>-3.802422</td>\n",
       "      <td>-2.534510</td>\n",
       "      <td>-1.783423</td>\n",
       "      <td>-1.594450</td>\n",
       "      <td>-0.753199</td>\n",
       "      <td>...</td>\n",
       "      <td>1.148884</td>\n",
       "      <td>0.958434</td>\n",
       "      <td>1.059025</td>\n",
       "      <td>1.371682</td>\n",
       "      <td>1.277392</td>\n",
       "      <td>0.960304</td>\n",
       "      <td>0.971020</td>\n",
       "      <td>1.614392</td>\n",
       "      <td>1.421456</td>\n",
       "      <td>1.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 141 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "        0         1         2         3         4         5         6    \\\n",
       "0 -0.112522 -2.827204 -3.773897 -4.349751 -4.376041 -3.474986 -2.181408   \n",
       "1 -1.100878 -3.996840 -4.285843 -4.506579 -4.022377 -3.234368 -1.566126   \n",
       "2 -0.567088 -2.593450 -3.874230 -4.584095 -4.187449 -3.151462 -1.742940   \n",
       "3  0.490473 -1.914407 -3.616364 -4.318823 -4.268016 -3.881110 -2.993280   \n",
       "4  0.800232 -0.874252 -2.384761 -3.973292 -4.338224 -3.802422 -2.534510   \n",
       "\n",
       "        7         8         9    ...       131       132       133       134  \\\n",
       "0 -1.818286 -1.250522 -0.477492  ...  0.792168  0.933541  0.796958  0.578621   \n",
       "1 -0.992258 -0.754680  0.042321  ...  0.538356  0.656881  0.787490  0.724046   \n",
       "2 -1.490659 -1.183580 -0.394229  ...  0.886073  0.531452  0.311377 -0.021919   \n",
       "3 -1.671131 -1.333884 -0.965629  ...  0.350816  0.499111  0.600345  0.842069   \n",
       "4 -1.783423 -1.594450 -0.753199  ...  1.148884  0.958434  1.059025  1.371682   \n",
       "\n",
       "        135       136       137       138       139  140  \n",
       "0  0.257740  0.228077  0.123431  0.925286  0.193137  1.0  \n",
       "1  0.555784  0.476333  0.773820  1.119621 -1.436250  1.0  \n",
       "2 -0.713683 -0.532197  0.321097  0.904227 -0.421797  1.0  \n",
       "3  0.952074  0.990133  1.086798  1.403011 -0.383564  1.0  \n",
       "4  1.277392  0.960304  0.971020  1.614392  1.421456  1.0  \n",
       "\n",
       "[5 rows x 141 columns]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#importing libraries and dataset\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import matplotlib.pyplot as plt\n",
    "from sklearn.metrics import accuracy_score\n",
    "from tensorflow.keras.optimizers import Adam\n",
    "from sklearn.preprocessing import MinMaxScaler\n",
    "from tensorflow.keras import Model, Sequential\n",
    "from tensorflow.keras.layers import Dense, Dropout\n",
    "from sklearn.model_selection import train_test_split\n",
    "from tensorflow.keras.losses import MeanSquaredLogarithmicError\n",
    "\n",
    "PATH_TO_DATA = 'http://storage.googleapis.com/download.tensorflow.org/data/ecg.csv'\n",
    "data = pd.read_csv(PATH_TO_DATA, header=None)\n",
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Z7txZa5WBPnc",
    "outputId": "7d8ac3b3-21a4-4be0-97ff-fc59b9750a97"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(4998, 141)"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#finding shape of the dataset\n",
    "data.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "id": "tytSKg7SDg6V"
   },
   "outputs": [],
   "source": [
    "#splitting training and testing dataset\n",
    "features = data.drop(140, axis=1)\n",
    "target = data[140]\n",
    "x_train, x_test, y_train, y_test = train_test_split(\n",
    "    features, target, test_size=0.2, stratify=target\n",
    ")\n",
    "train_index = y_train[y_train == 1].index\n",
    "train_data = x_train.loc[train_index]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "rCfSvzQoBZoy"
   },
   "outputs": [],
   "source": [
    "#scaling the data using MinMaxScaler\n",
    "min_max_scaler = MinMaxScaler(feature_range=(0, 1))\n",
    "x_train_scaled = min_max_scaler.fit_transform(train_data.copy())\n",
    "x_test_scaled = min_max_scaler.transform(x_test.copy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "id": "oIgBjiJHDpwl"
   },
   "outputs": [],
   "source": [
    "#creating autoencoder subclass by extending Model class from keras\n",
    "class AutoEncoder(Model):\n",
    "  def __init__(self, output_units, ldim=8):\n",
    "    super().__init__()\n",
    "    self.encoder = Sequential([\n",
    "      Dense(64, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(32, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(16, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(ldim, activation='relu')\n",
    "    ])\n",
    "    self.decoder = Sequential([\n",
    "      Dense(16, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(32, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(64, activation='relu'),\n",
    "      Dropout(0.1),\n",
    "      Dense(output_units, activation='sigmoid')\n",
    "    ])\n",
    "  \n",
    "  def call(self, inputs):\n",
    "    encoded = self.encoder(inputs)\n",
    "    decoded = self.decoder(encoded)\n",
    "    return decoded"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "aWbkLX8yB4EE",
    "outputId": "685dc961-4609-4d57-bdd2-c53e8f0ffed8"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "5/5 [==============================] - 1s 36ms/step - loss: 0.0112 - mse: 0.0252 - val_loss: 0.0134 - val_mse: 0.0309\n",
      "Epoch 2/20\n",
      "5/5 [==============================] - 0s 12ms/step - loss: 0.0106 - mse: 0.0238 - val_loss: 0.0129 - val_mse: 0.0297\n",
      "Epoch 3/20\n",
      "5/5 [==============================] - 0s 11ms/step - loss: 0.0095 - mse: 0.0214 - val_loss: 0.0125 - val_mse: 0.0289\n",
      "Epoch 4/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0085 - mse: 0.0190 - val_loss: 0.0123 - val_mse: 0.0283\n",
      "Epoch 5/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0075 - mse: 0.0168 - val_loss: 0.0120 - val_mse: 0.0277\n",
      "Epoch 6/20\n",
      "5/5 [==============================] - 0s 11ms/step - loss: 0.0068 - mse: 0.0152 - val_loss: 0.0119 - val_mse: 0.0275\n",
      "Epoch 7/20\n",
      "5/5 [==============================] - 0s 12ms/step - loss: 0.0062 - mse: 0.0139 - val_loss: 0.0116 - val_mse: 0.0269\n",
      "Epoch 8/20\n",
      "5/5 [==============================] - 0s 11ms/step - loss: 0.0058 - mse: 0.0129 - val_loss: 0.0114 - val_mse: 0.0262\n",
      "Epoch 9/20\n",
      "5/5 [==============================] - 0s 10ms/step - loss: 0.0055 - mse: 0.0122 - val_loss: 0.0111 - val_mse: 0.0255\n",
      "Epoch 10/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0052 - mse: 0.0117 - val_loss: 0.0107 - val_mse: 0.0248\n",
      "Epoch 11/20\n",
      "5/5 [==============================] - 0s 8ms/step - loss: 0.0050 - mse: 0.0112 - val_loss: 0.0105 - val_mse: 0.0242\n",
      "Epoch 12/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0049 - mse: 0.0109 - val_loss: 0.0103 - val_mse: 0.0238\n",
      "Epoch 13/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0048 - mse: 0.0107 - val_loss: 0.0101 - val_mse: 0.0234\n",
      "Epoch 14/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0047 - mse: 0.0105 - val_loss: 0.0101 - val_mse: 0.0233\n",
      "Epoch 15/20\n",
      "5/5 [==============================] - 0s 8ms/step - loss: 0.0046 - mse: 0.0104 - val_loss: 0.0100 - val_mse: 0.0233\n",
      "Epoch 16/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0046 - mse: 0.0103 - val_loss: 0.0100 - val_mse: 0.0232\n",
      "Epoch 17/20\n",
      "5/5 [==============================] - 0s 10ms/step - loss: 0.0046 - mse: 0.0102 - val_loss: 0.0100 - val_mse: 0.0231\n",
      "Epoch 18/20\n",
      "5/5 [==============================] - 0s 8ms/step - loss: 0.0045 - mse: 0.0102 - val_loss: 0.0100 - val_mse: 0.0231\n",
      "Epoch 19/20\n",
      "5/5 [==============================] - 0s 8ms/step - loss: 0.0045 - mse: 0.0101 - val_loss: 0.0100 - val_mse: 0.0231\n",
      "Epoch 20/20\n",
      "5/5 [==============================] - 0s 9ms/step - loss: 0.0045 - mse: 0.0100 - val_loss: 0.0099 - val_mse: 0.0229\n"
     ]
    }
   ],
   "source": [
    "#model configuration\n",
    "model = AutoEncoder(output_units=x_train_scaled.shape[1])\n",
    "model.compile(loss='msle', metrics=['mse'], optimizer='adam')\n",
    "epochs = 20\n",
    "\n",
    "history = model.fit(\n",
    "    x_train_scaled,\n",
    "    x_train_scaled,\n",
    "    epochs=epochs,\n",
    "    batch_size=512,\n",
    "    validation_data=(x_test_scaled, x_test_scaled)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 279
    },
    "id": "txWdeBzwEPIR",
    "outputId": "0393d6fc-fa45-442e-f16a-2a8867aae2b4"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAEGCAYAAABy53LJAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAAAykUlEQVR4nO3dd3xW5f3/8dcnO2QwAoSwN8jeoCJuUcpwC+Is1aJ1fqvVtj9b29rWuuoo1TpQbB2gUsWtdSHIRvYSkBFmmAmE7Ov3x7nBGDJJ7pw7yfv5eJzHfXLGfX/uY+Sdc51zrsucc4iIiJRXmN8FiIhIzaLgEBGRClFwiIhIhSg4RESkQhQcIiJSIRF+F1AdGjdu7Nq2bet3GSIiNcqiRYv2OOeaFF1eJ4Kjbdu2LFy40O8yRERqFDPbXNxyNVWJiEiFKDhERKRCFBwiIlIhdeIah4jUTbm5uaSmppKVleV3KSEtJiaGli1bEhkZWa7tFRwiUmulpqaSkJBA27ZtMTO/ywlJzjn27t1Lamoq7dq1K9c+aqoSkVorKyuLpKQkhUYpzIykpKQKnZUpOESkVlNolK2ix0jBUZpNs2HOJCgo8LsSEZGQoeAozcrp8PFv4OXRcGCr39WISA0THx/vdwlBoeAozYhHYNSTsG0xPH0qLJ0KGvhKROo4BUdpzKD/tXDTLGjaFf57I7xxHWTu87syEalBnHPcfffd9OjRg549ezJ16lQAduzYwbBhw+jTpw89evTg66+/Jj8/n+uuu+7Ytn//+999rv54uh23PBq1h+s/hNmPwxd/gS1z4cJJ0PEcvysTkXL6w7srWbU9vUrfs1vzRH4/qnuZ202fPp0lS5awdOlS9uzZw8CBAxk2bBivvvoqw4cP57e//S35+flkZmayZMkStm3bxooVKwA4cOBAldZcFXTGUV5h4XDaL+GGzyG2AfznEnj/LsjJ9LsyEQlxs2bNYty4cYSHh5OcnMzpp5/OggULGDhwIC+++CL3338/y5cvJyEhgfbt27Nx40ZuvfVWPvroIxITE/0u/zg646iolN5w45fw2R9h7j9h4xdw8bPQor/flYlIKcpzZhAsroRro8OGDWPmzJm8//77XH311dx9991cc801LF26lI8//phJkyYxbdo0Jk+eXM0Vl05nHCciMhbO/ytc8w7kHoHnz4Uv/wb5eX5XJiIhaNiwYUydOpX8/HzS0tKYOXMmgwYNYvPmzTRt2pQbbriBCRMmsHjxYvbs2UNBQQGXXHIJf/rTn1i8eLHf5R9HZxyV0f4MuOkb+OBu+PIv8N3HcNGz0Lij35WJSAi56KKLmDNnDr1798bMeOihh2jWrBlTpkzh4YcfJjIykvj4eF5++WW2bdvG9ddfT0Hg+bG//vWvPld/PCvpFKo2GTBggAv6QE4r3oL3/g/ysmH4AzBggndXloj4ZvXq1Zx00kl+l1EjFHeszGyRc25A0W3VVFVVelwCN8+BNifD+7+EVy6FjJ1+VyUiUuUUHFUpsTlcNd17cHDTLPjnEFj2BhTk+12ZiEiVUXBUNTMYdAP8/Gto2Bam/wye6g/z/gXZh/yuTkSk0hQcwdKkM0z4H1z2EsQ1hg9/BX/vBp/+Dg6m+l2diMgJU3AEU3gEdL8IfvY/mPAptD8TvnkKHu8Fb07w+sASEalhdDtudWk1yJv2b4b5z8KiKbDiTWh9Mpz8C+gywns6XUQkxOmMo7o1bAPD/wz/twqG/xXSt8HUq+CpfroOIiI1goLDLzGJcPLNcOu3cNkUiE/2roM81g0+uU/XQUTqoNLG79i0aRM9evSoxmpKpuDwW3gEdL8QJnziXUzveJY36uDjveDNn8LWBRoDRERCiq5xhJJWA6HVS3Bgi9dstfhl74n0pE7Q6wrodbnX1CUiFffhvbBzedW+Z7OecMGDJa6+5557aNOmDTfffDMA999/P2bGzJkz2b9/P7m5uTzwwAOMGTOmQh+blZXFTTfdxMKFC4mIiOCxxx7jzDPPZOXKlVx//fXk5ORQUFDAW2+9RfPmzbn88stJTU0lPz+f++67jyuuuKJSX1vBEYoatPaug5xxL6z8Lyx9Hb54wJtanwK9r4BuF3rdu4tIyBo7dix33HHHseCYNm0aH330EXfeeSeJiYns2bOHIUOGMHr0aKwCXRRNmjQJgOXLl7NmzRrOO+881q1bxzPPPMPtt9/O+PHjycnJIT8/nw8++IDmzZvz/vvvA3Dw4MFKfy8FRyiLToB+13jT/s2wfJo3fO27t8MHv4Iu50Ovsd6AUhFRflcrEtpKOTMIlr59+7J79262b99OWloaDRs2JCUlhTvvvJOZM2cSFhbGtm3b2LVrF82aNSv3+86aNYtbb70VgK5du9KmTRvWrVvHySefzJ///GdSU1O5+OKL6dSpEz179uSuu+7innvuYeTIkZx22mmV/l66xlFTNGwDw+6GWxZ4g0n1v87r1uT1cfBoF6+H3tRFuh4iEmIuvfRS3nzzTaZOncrYsWN55ZVXSEtLY9GiRSxZsoTk5GSysrIq9J4ldU575ZVXMmPGDGJjYxk+fDiff/45nTt3ZtGiRfTs2ZNf//rX/PGPf6z0d9IZR01j5g0a1aK/15y1/jNY9rr3XMj8ZyGpo3cWoushIiFh7Nix3HDDDezZs4evvvqKadOm0bRpUyIjI/niiy/YvHlzhd9z2LBhvPLKK5x11lmsW7eOLVu20KVLFzZu3Ej79u257bbb2LhxI8uWLaNr1640atSIq666ivj4eF566aVKfycFR00WHuk1V3U5H7IOwqp3vKasY9dDTobeY72n12Pq+12tSJ3UvXt3MjIyaNGiBSkpKYwfP55Ro0YxYMAA+vTpQ9euXSv8njfffDMTJ06kZ8+eRERE8NJLLxEdHc3UqVP5z3/+Q2RkJM2aNeN3v/sdCxYs4O677yYsLIzIyEiefvrpSn8njcdRGx3YAsumwbKpsGcdRMTCSaOg73hoOwzC1EIpdYPG4yi/iozHoTOO2qhBaxh2F5z2S68/rCWveN2bLJ8G9VtB73HQ50po1M7vSkWkBlJw1GZm0LK/Nw3/C6x5zwuRmQ/DzIegzVAvQLqNgeiSn1gVkeqzfPlyrr766h8ti46OZt68eT5VdLygBoeZnQ88AYQDzzvnHiyy3gLrRwCZwHXOucWBdZOBkcBu51yPQvs8DIwCcoANwPXOuQPB/B61QmQM9LzUmw6mes+GLHkF3rnZuyOr+0VeiLQ5RUPeSq3inKvQMxJ+69mzJ0uWLKnWz6zoJYugNXabWTgwCbgA6AaMM7NuRTa7AOgUmG4ECl+1eQk4v5i3/hTo4ZzrBawDfl21ldcB9Vt6TVm3Loaffgw9LoZVb8NLI+DJvvDVw3Bgq99VilRaTEwMe/furfA/jHWJc469e/cSExNT7n2CecYxCFjvnNsIYGavA2OAVYW2GQO87Lz/qnPNrIGZpTjndjjnZppZ26Jv6pz7pNCPc4FLg/YNajszaD3Emy74G6ya4Z2FfPEAfPFnaH86dL8YmnSFRu29Aalq0F9uIi1btiQ1NZW0tDS/SwlpMTExtGzZstzbBzM4WgCF/2xNBQaXY5sWwI5yfsZPgaknWqAUEhUHfcZ50/5NsOQ1WPIqvHvbD9tEJ3oX1Bu1h0YdAq/tIakDxDVRqEjIiYyMpF073QRS1YIZHMX9K1L0fLE82xT/5ma/BfKAV0pYfyNe8xetW7cuz1vKUQ3bwpm/htPvgX0bA9OGH+a3L/HOTlz+D/tExQdCpUigNO4CcUl+fRMRCYJgBkcq0KrQzy2B7SewzXHM7Fq8C+dnuxIaL51zzwLPgvccR/nLlmPCwqBxR28qKj/Xe17kaJjsDQTLzuXe3VsFed52FgbtTveeZO860huHRERqtGAGxwKgk5m1A7YBY4Eri2wzA7glcP1jMHDQOVdqM1XgTq17gNOdc5lVX7aUS3ikd0aR1OH4dfm5cHAr7N0IW+bA8jfg7Zsg4k5viNxel0OHs9Uxo0gNFdQnx81sBPA43u24k51zfzaziQDOuWcCt+P+A+/uqUy8W2sXBvZ9DTgDaAzsAn7vnHvBzNYD0cDewMfMdc5NLK2OOvfkeKhxDrbO9x5AXDEdjuyD2Ebe3Vw9L/fGYtf1EZGQU9KT4+pyRKpXXg5s+NzrDmXtB5CX5V1T6XmZFyJNOvtdoYgEKDgUHKEnOwNWv+eFyPdfgSuAlD7eaIc9LoGEZL8rFKnTFBwKjtCWsdMbJnfZNNixxLuo3v4ML0ROGg1R9fyuUKTOUXAoOGqOtHXe9ZBl0+DAZoiu711Q73+tN8aziFQLBYeCo+ZxDjbP9gapWvUO5GdD835egPS4xBtaV0SCRsGh4KjZMvd5ZyCLp8DuVd4Dhz0u9obQbd5Pd2WJBIGCQ8FROzgHqQth0UuwcjrkZkJyT+8spOdlENvA7wpFag0Fh4Kj9slK9x4uXDwFdiz1RjrsfiH0u9bruFFnISKVouBQcNRu25d4AbLsDcjJ8PrI6neNN9qh+soSOSEKDgVH3ZBz2Hs6ffEUSF0A4dHe4FWDfw4pvf2uTqRGUXAoOOqeXatgwfPeaIe5h6H1yV6AdB0F4Ro1WaQsCg4FR9115AB8+x+Y/6z3XEhiCxg4Afpdp2YskVIoOE4gOPILHEdy84mP1l+ntUJBPqz7GOY943VxEhHj3Yk1+Od6sFCkGCUFR9DGHK8NHvpoDRf/czZb96n39lohLBy6joBrZ8DNc70L58vfhGeGwos/8Qanys/zu0qRkKfgKMWwzk3YeTCLi/45m0Wb9/tdjlSlpifBqMfh/1bBuX/yBqWadjU82QdmPe49cCgixVJTVRnW7z7EhCkL2HEwi0cu683o3s2ruDoJCQX5sPZDrxlr09feMyG9LveasZK7+12diC90jaMSF8f3Hc7h5/9eyIJN+7nznM7cdnZHTA+X1V47V8D8f3ldnORleUPfnvwL6HiuN5yuSB2h4KjkXVXZefn8+q3lTP92Gxf2ac6Dl/QiJjK8iiqUkJS5Dxa9CPOfg4wdkNQJhkz0ro1ExfldnUjQKTiq4HZc5xyTvljPI5+sY0Cbhvzr6v4kxUdXQYUS0vJyYNXbMGeSN1ZITAMYcD0MuhES1XQptZeCowqf43hv2XZ+OW0pyYkxTL5uAB2bqnvvOsE52DIX5k6CNe97g011uxBOvhla9Pe7OpEqp+Co4gcAv92ynxteXkh2XgFPj+/P0E6Nq/T9JcTt3wTz/gWL/+31jdVqiBcgXUd6t/2K1AIKjiA8OZ66P5MJLy1kfdoh/jSmB1cObl3lnyEhLivdeyp93jPeU+kNWsPgidD3aohJ9Ls6kUpRcASpy5GMrFxufe1bvlybxs+GtuPXI04iPEx3XNU5Bfle89Xcf8KWORCVAH2v8m7nbdTO7+pEToiCI4h9VeXlF/DA+6t56ZtNnHNSMk+M7UOcuimpu7YthrlPewNNFeR5nSv2vNS7HhKnJk2pORQc1dDJ4ZRvNvGHd1fStVkiL1w3gJT6sUH/TAlh6dvh21dgxZuQtgYsHDqcCT0uha4/UVOWhDwFRzX1jvvF2t3c+uq3xEWH8/w1A+nZsn61fK6EMOdg10pvtMIV0+HgFq+Dxc7DvRDpdB5ExvhdpchxFBzV2K362p0Z/PSlBew7nMPjY/swvHuzavtsCXHOwdb53lnIiumQuQeiE+GkUdDjEu8pdY0VIiFCwVHN43GkZWRzw8sLWbn9IO/8YijdmqtZQorIz/O6d1/xFqx+F7LTIa4JdL/IOxNpNUjjpouvFBw+DOS073AO5/19Jo3jo5hxy1CiItTPkZQgNwvWf+o1Z6372Osjq35r6HERdBruhUh4pN9VSh2j4PBpBMD/rdrFz15eyC/O7MDdw7v6UoPUMFnpsPYDb6yQjV94d2ZFJ0L706HjOd5Uv6XfVUodUFJwqDE1yM7plsxl/Vvy9JcbOOekZPq2buh3SRLqYhKh91hvyjoIG7+C9f/zptXvets0OQk6nu2FSJtTIEJ9pkn10RlHNUjPyuWCx78mOiKM9287jdgodUkhJ8A577beoyGy+RvIz4HIetBuWOBs5Gxo1N7vSqWWUFOVj8EBMHv9HsY/P4/rT23L70dpYCCpAjmH4fuvA0Hyqdd/FnjB0fEcb/yQtkMhqp6vZUrNpaYqn53asTHXntyGF2dv4txuyZzSQU8QSyVFxUGX870JYO+GH85GFv8b5j8LYZGQ0gtaDgxMA6BBG92tJZWiM45qlJmTx4gnviY33/HRHaeREKO7ZCRIcrNgyzew8UtIXQTbF0Nuprcurgm0GOCFSMuB0KIfRGtoADmemqpCIDgAFm3ex2XPzOHyAa148JJefpcjdUV+HuxeBakLIHWh97r3u8BKg6bdfgiSlgOhcWcNkytqqgoV/ds04sZhHXjmqw2c1z2Zs7om+12S1AXhEV6TVUovGDjBW5a5z+uQcVsgSFa9DYuneOuiE70zkZQ+EJ8MsQ2hXiOIbRR4bQgx9TX2SB0V1DMOMzsfeAIIB553zj1YZL0F1o8AMoHrnHOLA+smAyOB3c65HoX2aQRMBdoCm4DLnXP7S6sjlM44wBu/fPRTs9mXmcMndwyjYVyU3yWJQEEB7NsQOCsJnJnsWgkuv4QdDGIbeCFSOFB+NN/Q65crMsZ7jYgtNB8DkbHercQRsTrDCUHV3lRlZuHAOuBcIBVYAIxzzq0qtM0I4Fa84BgMPOGcGxxYNww4BLxcJDgeAvY55x40s3uBhs65e0qrJdSCA2DFtoNcOGk2F/RM4alxff0uR6R4BQWQfdA7Ozmy35sy98GRfaXMH/C6T6mosMhAkMT8OGyiE71rMMdNxS0vtCwyVjcBVJIfTVWDgPXOuY2BAl4HxgCrCm0zBi8YHDDXzBqYWYpzbodzbqaZtS3mfccAZwTmpwBfAqUGRyjq0aI+t5/diUc/Xcfw7smM7NXc75JEjhcW9sOZQ0Xk53oBcmQ/5B2BvGzIPeJ1pZKX5V28L8/y3COQcwjSUyE7w5uy0qEgt+waLPyHMIlJLDIf+PnofEz94tdHJ3rNcQqgH6lQcJhZGBDvnCvPnxMtgK2Ffk7FO6soa5sWwI5S3jfZObcDwDm3w8yallDrjcCNAK1bh+aQrjed0YH/rd7FfW+vYFC7RjRNUNfaUkuER0J8E28KhrzsQJCk/xAox6b0H4dM4WUZO2DP2ooF0FEWVspkpa8PC/eC7NhrhBfKR5eFRQTmiywLi/Dugkto5l1rKvwa19S3npTL/FQzexWYCOQDi4D6ZvaYc+7hsnYtZlnRdrHybHNCnHPPAs+C11RVFe9Z1SLCw3j08t6MeHIWv35rOc9fOwDTXzYiZYuI9qbKjKjonHeWcyxcDv44aLLSvbOdgnzAgSsoZippeWAqODqf7/U5VpAfmC/wfnb5hZblQ17OD/MFed6Z25Y5kLm3mC9g3vePbwYJyT+8JqQUCZkUiKja66jliatuzrl0MxsPfIDXLLQIKCs4UoFWhX5uCWw/gW2K2nW0OcvMUoDdZX2BUNaxaQK/Gt6FB95fzRuLUrl8QKuydxKRyjPzroNExnr/4IayvBw4vBsydsGhnZCxEw7t+vHrrpVwaPfxNzOMex26XFCl5ZQnOCLNLBK4EPiHcy7XzMrzF/wCoJOZtQO2AWOBK4tsMwO4JXD9YzBw8GgzVClmANcCDwZe3ylHLSHtp6e245NVu/jju6s4pUMSLRuqiwgRKSQiyusRuaxekQvyvbOTwoGS0qfKyynP/W//wrvtNQ6YaWZtgDKvcTjn8oBbgI+B1cA059xKM5toZhMDm30AbATWA88BNx/d38xeA+YAXcws1cwCN5/zIHCumX2Hd8fWj27xrYnCwoxHL+tNgXP86s1lFBSEZMuaiIS6sHCIb+o9r9PpXOh3NSSmVPnHnNDtuGYWEQiGGiEUb8ctzqvztvCb/y7nD6O7c+0pbf0uR0TquJJuxy3zjMPMbjezRPO8YGaLgbOCUmUdN25QK07v3IS/fria7/cc9rscEZFilaep6qeB22/PA5oA11MLmodCkZnxt0t6ERUexi+nLSFfTVYiEoLKExxH7w8dAbzonFtK8bfRShVoVj+GP13Yg8VbDvDszI1+lyMicpzyBMciM/sELzg+NrMEoCC4ZdVto3s354Iezfj7p+tYs/MEum4QEQmi8gTHBOBeYKBzLhOIwmuukiAxMx64sAeJsRHcOXUpOXnKaREJHWUGh3OuAO/BvP9nZo8ApzjnlgW9sjouKT6av17ci9U70nn007V+lyMickx57qp6ELgdr3PCVcBtZvbXYBcmcG63ZMYNas2zMzfyzYY9fpcjIgKUr6lqBHCuc26yc24ycD7wk+CWJUfdN/Ik2iXF8ctpSzmYWYEO2UREgqS8I6c0KDRfPwh1SAnqRUXw+Ng+pGVk85u3l1MXhvoVkdBWnuD4K/Ctmb1kZlPwOjj8S3DLksJ6tWzAned25v1lO5i+eJvf5YhIHVeei+OvAUOA6YHpZOD7INclRUw8vQOD2jbi9zNWsnVfpt/liEgdVq6mqsCIfDOcc+8453YCbwS5LikiPMx47IreGHDH1CXk5esWXRHxx4mODq8nx33QsmE9HrioB4s27+efX27wuxwRqaNONDh0hdYnY/q0YEyf5jzx2Xd8u2W/3+WISB1U4kBOZvYuxQeEAUlBq0jK9McxPVi4aT93TF3CB7edRly0P+MOi0jdVNq/OI+c4DoJsvqxkTx2eW/GPjeXP7y7kocu7e13SSJSh5QYHM65r6qzEKmYwe2TuOn0Dvzzyw2c1bUp5/eo+lG+RESKc6LXOCQE3HFOZ3q2qM+905ez82CW3+WISB2h4KjBoiLCeHxsH7JzC7jrjaUaq1xEqsUJBYeZ6WpsiOjQJJ7/N/IkZq3fw+TZei5TRIKvxOAws1mF5v9dZPX8oFUkFXbloNacc1IyD320ltU7NPCTiARXaWcccYXmuxdZpwcAQ4g3VnlP6teL5I7Xl5CVm+93SSJSi5UWHKU1mKsxPcQkxUfz8KW9WLsrg799tMbvckSkFivtWkUDM7sIL1wamNnFgeWGulYPSWd0acp1p7TlxdmbOKNLU07v3MTvkkSkFirtjOMrYDQwMjA/KjCNBGYGvzQ5Efde0JXOyfHc9cZS9h7K9rscEamFSnsA8PqS1pnZJcEpRyorJjKcx6/oy4WTZnPv9OU8e3V/zHRJSkSqzok+x/H3Kq1CqlS35oncPbwLn67axesLtvpdjojUMupWvZaaMLQdp3ZM4o/vrmJD2iG/yxGRWkTdqtdSYWHGY5f3ISYyjFtf/ZbsPN2iKyJVo7QHAJeb2bJipuVAcjXWKCcoOTGGhy/tzaod6Tz00Vq/yxGRWqK023FHVlsVEjTndEvmmpPb8MKs7zmtU2PO6NLU75JEpIYr8YzDObe58AQcAvoBjQM/Sw3xmxEn0SU5gbveWMruDPWiKyKVU1pT1Xtm1iMwnwKsAH4K/NvM7qie8qQqxESG89SVfcnIyuOX09SLrohUTmkXx9s551YE5q8HPnXOjQIG4wWI1CCdkxO4b2Q3vv5uDy/MUi+6InLiSguO3ELzZwMfADjnMoCCYBYlwTF+cGvO65bMQx+vYXnqQb/LEZEaqrTg2Gpmtwb6q+oHfARgZrFAZHne3MzON7O1ZrbezO4tZr2Z2ZOB9cvMrF9Z+5pZHzOba2ZLzGyhmQ0q75et67xedHuRFBfNba9/y+HsPL9LEpEaqLTgmIDXnfp1wBXOuQOB5UOAF8t6YzMLByYBFwDdgHFm1q3IZhcAnQLTjcDT5dj3IeAPzrk+wO8CP0s5NYyL4vGxfdi09zC/n7HS73JEpAYq7a6q3c65ic65Mc65Twot/8I590g53nsQsN45t9E5lwO8Dowpss0Y4GXnmYvXC29KGfs6IDEwXx/YXo5apJAh7ZO45cyOvLkolRlLdfhEpGJKfI7DzGaUtqNzbnQZ790CKNxRUirehfWytmlRxr53AB+b2SN4wXdKGXVIMW4/uxOz1+/ht9OX07dVA1o1qud3SSJSQ5TWVHUy0BL4GngEeLTIVJbi+rMqeh9oSduUtu9NwJ3OuVbAncALxX642Y2BayAL09LSylFu3RIRHsYTY/sCcNvr35Kbr/sdRKR8SguOZsBvgB7AE8C5wB7n3FfOua/K8d6pQKtCP7fk+GalkrYpbd9rgemB+TfwmrWO45x71jk3wDk3oEkTDWhUnFaN6vHni3vy7ZYDPPG/7/wuR0RqiNKuceQ75z5yzl2Ld0F8PfClmd1azvdeAHQys3ZmFgWMBYo2f80ArgncXTUEOOic21HGvtuB0wPzZwH6F68SRvduzmX9WzLpy/XM2bDX73JEpAYora8qzCwa+AkwDmgLPMkPf+2XyjmXZ2a3AB8D4cBk59xKM5sYWP8M3rMhI/BCKRPvQcMS9w289Q3AE2YWAWTh3Y0llXD/6O4s2ryfO6cu4cPbT6NhXJTfJYlICDPniu9+wsym4DVTfQi8Xugp8hpnwIABbuHChX6XEdJWbDvIRf+czZldmvIvjRooIoCZLXLODSi6vLRrHFcDnYHbgW/MLD0wZZhZerAKFX/0aFGfXw3vyierdvHKvC1+lyMiIay0McdPdJAnqaEmDG3H1+v38Kf3VjGwbSO6NEvwuyQRCUEKBzkmLMx49LLeJMREcNtr35KVq1EDReR4Cg75kSYJ0TxyWW/W7srgLx+s9rscEQlBCg45zhldmjJhaDtenrOZT1ft8rscEQkxCg4p1q/O70L35onc9cZSVu/QvRAi8gMFhxQrOiKcZ67qT2xkOOOfn8e6XRl+lyQiIULBISVq1ager904hIgw48rn5rF+9yG/SxKREKDgkFK1axzHqzcMARxXPjeX7/cc9rskEfGZgkPK1LFpPK/eMIS8Ai88tuzN9LskEfGRgkPKpXNyAv+ZMJgjufmMe24uW/cpPETqKgWHlFu35on8Z8JgMrJyufL5uWw/cMTvkkTEBwoOqZAeLerz7wmDOXA4l3HPzWXnwSy/SxKRaqbgkArr3aoBUyYMYu+hHK58bi670xUeInWJgkNOSL/WDXnp+oHsTM/iyufnkZaR7XdJIlJNFBxywga0bcTk6waSuj+Tq56fx77DOX6XJCLVQMEhlTKkfRKTrx3Ipr2HGf/8PA5kKjxEajsFh1TaKR0b89w1A9iQdoirXpjHwcxcv0sSkSBScEiVGNa5Cf+6qj/rdh7imsnzSM9SeIjUVgoOqTJndm3KP8f3Y+X2dK6bPJ9D2Xl+lyQiQaDgkCp1Trdk/nFlX5amHuT6F+dzWOEhUusoOKTKnd8jhcev6MOizfu5/qUF7NfdViK1ioJDgmJU7+Y8PrYvS7YcYNQ/ZrFi20G/SxKRKqLgkKAZ3bs5b0w8mYICx8VPf8O0hVv9LklEqoCCQ4Kqd6sGvHvrUAa2bciv3lzGr6cvJzsv3++yRKQSFBwSdEnx0bz808HcfEYHXpu/hcufmaOedUVqMAWHVIvwMONX53flmav6syHtMCOfmsXs9Xv8LktEToCCQ6rV+T2aMeOWU0mKi+LqF+bx9JcbcM75XZaIVICCQ6pd+ybxvP2LUxnRM4W/fbSGif9ZRIaeNBepMRQc4ou46AieGteX+0Z243+rdzPmH7P5bleG32WJSDkoOMQ3ZsaEoe149WeDSc/KY8yk2by3bLvfZYlIGRQc4rvB7ZN4/7ahnJSSyC2vfsuf3ltFbn6B32WJSAkUHBISkhNjeO2GIVx3SltemPU945+fx+4MDUkrEooUHBIyoiLCuH90dx6/og/LUg8w6qlZLNq8z++yRKQIBYeEnAv7tuC/N59KTGQ4lz0zh/veXqGOEkVCiIJDQtJJKYnMuGUoVw9pw6vzt3Dmo1/y7zmbyNO1DxHfBTU4zOx8M1trZuvN7N5i1puZPRlYv8zM+pVnXzO7NbBupZk9FMzvIP6pHxvJH8b04IPbTqNbSiL3vbOSkU/NYs6GvX6XJlKnBS04zCwcmARcAHQDxplZtyKbXQB0Ckw3Ak+Xta+ZnQmMAXo557oDjwTrO0ho6NIsgVd+Npinx/cjIyuPcc/N5eZXFpG6P9Pv0kTqpGCecQwC1jvnNjrncoDX8f7BL2wM8LLzzAUamFlKGfveBDzonMsGcM7tDuJ3kBBhZlzQM4XPfnk6/3duZz5fs5uzH/2Kxz5dx5Ec9bYrUp2CGRwtgMIDMKQGlpVnm9L27QycZmbzzOwrMxtY3Ieb2Y1mttDMFqalpVXia0goiYkM57azO/HZL8/g3G7JPPnZd5z96Je8t2y7+rwSqSbBDA4rZlnR/7NL2qa0fSOAhsAQ4G5gmpkdt71z7lnn3ADn3IAmTZqUv2qpEVo0iOUfV/Zj6o1DqF8vilte/Zaxz85l1fZ0v0sTqfWCGRypQKtCP7cEivYnUdI2pe2bCkwPNG/NBwqAxlVYt9Qgg9sn8d6tQ/nzRT1YtyuDkU99zW//u5x9un1XJGiCGRwLgE5m1s7MooCxwIwi28wArgncXTUEOOic21HGvm8DZwGYWWcgCtDADnVYeJgxfnAbvrzrTK45uS2vL9jKmY98yZRvdPuuSDAELTicc3nALcDHwGpgmnNupZlNNLOJgc0+ADYC64HngJtL2zewz2SgvZmtwLtofq1T47YA9etFcv/o7nxw22n0aJHI72es5Pwnvmbawq0arlakClld+Dd3wIABbuHChX6XIdXIOcfHK3fx90/XsXZXBo3jo7l6SBuuGtKapPhov8sTqRHMbJFzbsBxyxUcUps555i9fi8vzNrIF2vTiIoI4+K+LZgwtB2dkhP8Lk8kpJUUHBF+FCNSXcyMoZ0aM7RTY9bvzmDy7E28tSiV1xdsZVjnJvxsaDtO69SYYm7ME5ES6IxD6px9h3N4dd5mpszZTFpGNp2T45kwtB1j+rQgJjLc7/JEQoaaqhQcUkR2Xj7vLd3B87O+Z/WOdJLiorhqSBuuGtKGJgm6DiKi4FBwSAmcc8zZuJfJs77nf6t3ExUexpg+zZlwWju6Nkv0uzwR3+gah0gJzIxTOjTmlA6N2Zh2iBdnb+LNRam8sSiVUzokcXG/lgzvnkxCTKTfpYqEBJ1xiBTjQGYOr87fwmvzt7B13xGiIsI4q0tTRvdpzlldm+paiNQJaqpScMgJcM7x7dYDzFiynfeW7WDPoWzioyM4r3syo3s359SOjYkM13hoUjspOBQcUkn5BY45G/YyY+k2Plyxk4ysPBrFRfGTnimM7tOc/q0bEham23ql9lBwKDikCmXn5fPV2jTeWbqdz1bvIiu3gOb1YxjVuzmjejene/NEPRsiNZ6CQ8EhQXIoO4//rdrFO0u28fV3e8grcHRoEsfo3i0Y1TuF9k3i/S5R5IQoOBQcUg32Hc7hwxU7mLFkO/M37cM5aN84jrO6NuWsrk0Z0LYRURG6JiI1g4JDwSHVbMfBI3y0Yiefr9nNvI37yMkvICE6gtM6N+bMLk05s2tTGqvDRQlhCg4Fh/jocHYes9fv4fM1u/li7W52pWdjBr1aNuCsLk05+6Smui4iIUfBoeCQEOGcY+X2dD5fs5vP1+xmaeoBnIOmCdGc1dU7ExnasTFx0Xo+V/yl4FBwSIjacyibL9em8cWa3cxcl0ZGdh5R4WEMbt+I0zs3oUeL+nRtlkCDelF+lyp1jIJDwSE1QG5+AQs27eOLwNnIhrTDx9Y1S4yha0oCXZsl0rVZAl1TEmjfOF4X2yVoFBwKDqmBdqdnsXpnBmt2pLN2Zward2awfncGufne/7eR4UaHJvGBIAkESrNEkhOjdb1EKk2dHIrUQE0TY2iaGMPpnZscW5abX8DGtMOs2ZnOmkCozP9+H28v2X5smwb1IunaLIEuyQm0bRxH26Q4WifVo1XDejpDkUpTcIjUMJHhYXRplkCXZgmMKbT8YGYua3ams3ZXBqt3ZLBmZzpvLd7Goey8Y9uEGTRvEEubpHq0SYqjbVI9WjeKo23jerRuVI96UfonQcqm3xKRWqJ+vUgGt09icPukY8ucc+w9nMPmvYfZvDeTTXszj81/uHwH+zNzf/QeTROij52dtE2qR+ukOFLqx9AsMYamidFER6hXYFFwiNRqZkbj+Ggax0fTv02j49YfPJLLlr2ZbNp7mC37Mtm0xwuVr79L481F2cdtnxQXRXJiDCn1Y0iuH0NKovfarNCyhOgIXV+p5RQcInVY/dhIerasT8+W9Y9bl5mTR+r+I+w4mMWug1nsTM/y5gOv3249wL7DOcftFxcVfixMkhNjqB8bSWJMBImxkSTGRJIYG0FiTCQJP5qPIELd09cYCg4RKVa9qAg6JyfQOTmhxG2y8/LZnZ7NjkCw7Dr4Q7jsTM9iwaZ9pB/JJSM7j7Ju4IyLCv9RmCQGAqdBvSjqx0bSoN7RKYoGsT+8JsZGEq7u7KuVgkNETlh0RDitGtWjVaN6pW5XUOA4nJNHelYe6UdyvSkwn5H1w3x6Vi7pR/JIz8pld0YW3+3O5WCmt740iTERNIzzgqT+sWCJpEFsJHHREdSLjiAuKpx6URHER0dQLzqcuKgI6kWFH/s5KjxMTWzlpOAQkaALCzMSAs1TLRrEVnj//AJH+pFcDhzJ5UBmDgcyczlwJPCamcvBwPL9md42W/Ye5sARb3l5H1WLCLNCQVIoaGIiSIj2XuMLv0b/8HNCdOSPlsdE1u4QUnCISMgLDzMaxkXRMC4KiCv3fgUFjqy8fA5n53M4O4/DOXlk5uRzKDuPzOx87+fsPA7neOuPrcvJO7bP1n2ZHMrO86asPPIKyk6i8DD7UYhER4QTExlGTGQ4MZHhREccnffWRUeGERMR/qNlR7ePjvhhm6PzURFH58OIDmwTEWbVFlYKDhGptcLCjHpREdSLiqBJQuW7sHfOkZ1XQEZWHocDYZKRFQiV7FwOZeWRkR1YF5jPzi0gKzef7DzvNT0rl6zcArLz8sk6ui63gJz8gsp9V+PHoRIIoL9c1JNB7Y6/o64yFBwiIuVkZsfOGqoiiAorKHDHwiUrzwuTrLx8cvIKyM4rIDsQNtl53mtxy48tC+yfnVdAfBB6WVZwiIiEgLAwIzYqnNio0H/IUjdOi4hIhSg4RESkQhQcIiJSIQoOERGpEAWHiIhUiIJDREQqRMEhIiIVouAQEZEKMVfeHsBqMDNLAzaf4O6NgT1VWE5VU32Vo/oqR/VVXijX2MY516TowjoRHJVhZgudcwP8rqMkqq9yVF/lqL7Kqwk1FqWmKhERqRAFh4iIVIiCo2zP+l1AGVRf5ai+ylF9lVcTavwRXeMQEZEK0RmHiIhUiIJDREQqRMERYGbnm9laM1tvZvcWs97M7MnA+mVm1q8aa2tlZl+Y2WozW2lmtxezzRlmdtDMlgSm31VXfYHP32RmywOfvbCY9X4evy6FjssSM0s3szuKbFOtx8/MJpvZbjNbUWhZIzP71My+C7w2LGHfUn9Xg1jfw2a2JvDf779m1qCEfUv9XQhiffeb2bZC/w1HlLCvX8dvaqHaNpnZkhL2DfrxqzTnXJ2fgHBgA9AeiAKWAt2KbDMC+BAwYAgwrxrrSwH6BeYTgHXF1HcG8J6Px3AT0LiU9b4dv2L+W+/Ee7DJt+MHDAP6ASsKLXsIuDcwfy/wtxLqL/V3NYj1nQdEBOb/Vlx95fldCGJ99wN3leO/vy/Hr8j6R4Hf+XX8KjvpjMMzCFjvnNvonMsBXgfGFNlmDPCy88wFGphZSnUU55zb4ZxbHJjPAFYDLarjs6uQb8eviLOBDc65E+1JoEo452YC+4osHgNMCcxPAS4sZtfy/K4GpT7n3CfOubzAj3OBllX9ueVVwvErD9+O31FmZsDlwGtV/bnVRcHhaQFsLfRzKsf/w1yebYLOzNoCfYF5xaw+2cyWmtmHZta9eivDAZ+Y2SIzu7GY9SFx/ICxlPw/rJ/HDyDZObcDvD8WgKbFbBMqx/GneGeQxSnrdyGYbgk0pU0uoakvFI7facAu59x3Jaz38/iVi4LDY8UsK3qfcnm2CSoziwfeAu5wzqUXWb0Yr/mlN/AU8HZ11gac6pzrB1wA/MLMhhVZHwrHLwoYDbxRzGq/j195hcJx/C2QB7xSwiZl/S4Ey9NAB6APsAOvOago348fMI7Szzb8On7lpuDwpAKtCv3cEth+AtsEjZlF4oXGK8656UXXO+fSnXOHAvMfAJFm1ri66nPObQ+87gb+i9ckUJivxy/gAmCxc25X0RV+H7+AXUeb7wKvu4vZxu/fw2uBkcB4F2iQL6ocvwtB4Zzb5ZzLd84VAM+V8Ll+H78I4GJgaknb+HX8KkLB4VkAdDKzdoG/SscCM4psMwO4JnB30BDg4NFmhWALtIm+AKx2zj1WwjbNAtthZoPw/tvurab64sws4eg83kXUFUU28+34FVLiX3p+Hr9CZgDXBuavBd4pZpvy/K4GhZmdD9wDjHbOZZawTXl+F4JVX+FrZheV8Lm+Hb+Ac4A1zrnU4lb6efwqxO+r86Ey4d31sw7vjovfBpZNBCYG5g2YFFi/HBhQjbUNxTudXgYsCUwjitR3C7AS7y6RucAp1Vhf+8DnLg3UEFLHL/D59fCCoH6hZb4dP7wA2wHk4v0VPAFIAj4Dvgu8Ngps2xz4oLTf1Wqqbz3e9YGjv4PPFK2vpN+Faqrv34HfrWV4YZASSscvsPylo79zhbat9uNX2UldjoiISIWoqUpERCpEwSEiIhWi4BARkQpRcIiISIUoOEREpEIUHCKVYGb59uOed6ust1Uza1u4d1WRUBHhdwEiNdwR51wfv4sQqU464xAJgsCYCn8zs/mBqWNgeRsz+yzQEd9nZtY6sDw5MMbF0sB0SuCtws3sOfPGYfnEzGID299mZqsC7/O6T19T6igFh0jlxBZpqrqi0Lp059wg4B/A44Fl/8DrXr4XXieBTwaWPwl85bxOFvvhPTUM0AmY5JzrDhwALgksvxfoG3ificH5aiLF05PjIpVgZoecc/HFLN8EnOWc2xjooHKncy7JzPbgdYWRG1i+wznX2MzSgJbOuexC79EW+NQ51ynw8z1ApHPuATP7CDiE14vv2y7QQaNIddAZh0jwuBLmS9qmONmF5vP54brkT/D6/uoPLAr0uipSLRQcIsFzRaHXOYH5b/B6ZAUYD8wKzH8G3ARgZuFmlljSm5pZGNDKOfcF8CugAXDcWY9IsOivFJHKiTWzJYV+/sg5d/SW3Ggzm4f3B9q4wLLbgMlmdjeQBlwfWH478KyZTcA7s7gJr3fV4oQD/zGz+ni9Dv/dOXegir6PSJl0jUMkCALXOAY45/b4XYtIVVNTlYiIVIjOOEREpEJ0xiEiIhWi4BARkQpRcIiISIUoOEREpEIUHCIiUiH/HxAJA4STmJ4hAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.xlabel('Epochs')\n",
    "plt.ylabel('MSLE Loss')\n",
    "plt.legend(['loss', 'val_loss'])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "hwaxD_pZETdl",
    "outputId": "d3a4df47-0b0a-466a-87bb-aec715c37db7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "73/73 [==============================] - 0s 798us/step\n",
      "Threshold: 0.009784297004257442\n"
     ]
    }
   ],
   "source": [
    "#finding threshold for anomaly and doing predictions\n",
    "def find_threshold(model, x_train_scaled):\n",
    "  reconstructions = model.predict(x_train_scaled)\n",
    "  reconstruction_errors = tf.keras.losses.msle(reconstructions, x_train_scaled)\n",
    "  threshold = np.mean(reconstruction_errors.numpy()) \\\n",
    "   + np.std(reconstruction_errors.numpy())\n",
    "  return threshold\n",
    "\n",
    "def get_predictions(model, x_test_scaled, threshold):\n",
    "  predictions = model.predict(x_test_scaled)\n",
    "  errors = tf.keras.losses.msle(predictions, x_test_scaled)\n",
    "  anomaly_mask = pd.Series(errors) > threshold\n",
    "  preds = anomaly_mask.map(lambda x: 0.0 if x == True else 1.0)\n",
    "  return preds\n",
    "\n",
    "threshold = find_threshold(model, x_train_scaled)\n",
    "print(f\"Threshold: {threshold}\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ph9bwzz0CLtl",
    "outputId": "b6ef6aa1-c1c8-413e-fc37-c35b87784e67"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "32/32 [==============================] - 0s 672us/step\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "0.947"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#getting accuracy score\n",
    "predictions = get_predictions(model, x_test_scaled, threshold)\n",
    "accuracy_score(predictions, y_test)"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "collapsed_sections": [],
   "provenance": []
  },
  "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": "3.8.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}