LSTM_AE_IDS.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "provenance": [],
      "collapsed_sections": [],
      "mount_file_id": "1d_2T8v_kYeBRdI7jzaVLvnDb2VXG5fng",
      "authorship_tag": "ABX9TyNFZwZoY3IMaA/GSploWmSP",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/seoyeon7/ML_IDS_Project/blob/main/LSTM_AE_IDS.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ejOeL1MQW_NB"
      },
      "outputs": [],
      "source": [
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "%matplotlib inline\n",
        "\n",
        "import os\n",
        "os.environ['CUDA_VISIBLE_DEVICES']=''\n",
        "\n",
        "import pandas as pd\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "import joblib\n",
        "from sklearn import metrics\n",
        "\n",
        "import tensorflow as tf\n",
        "tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.ERROR)\n",
        "\n",
        "from tensorflow import keras\n",
        "from tensorflow.keras import layers\n",
        "from tensorflow.keras.models import load_model\n",
        "from tensorflow.keras.callbacks import EarlyStopping, ModelCheckpoint\n",
        "\n",
        "import random as rn\n",
        "from numpy.random import seed\n",
        "\n",
        "from scipy.io.arff import loadarff\n",
        "from sklearn.metrics import roc_curve, auc, precision_recall_curve"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 데이터 불러오기"
      ],
      "metadata": {
        "id": "6ZUir8itrWPV"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "noraml_data_raw = pd.read_csv(\"/content/normal.csv\")\n",
        "abnoraml_data_raw = pd.read_csv(\"/content/abnormal.csv\")"
      ],
      "metadata": {
        "id": "bPOIFnXeYhHk"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 데이터 리스트화\n",
        "normal_data_list = noraml_data_raw['ID'].values.tolist()\n",
        "abnormal_data_list = abnoraml_data_raw['ID'].values.tolist()"
      ],
      "metadata": {
        "id": "KLqNHUo4ZQTO"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 데스트 데이터, 검정 데이터 분리\n",
        "\n",
        "normal_data = np.array(normal_data_list).reshape(10485,100)\n",
        "normal_data_train = normal_data[:8000]\n",
        "normal_data_test = normal_data[8000:]\n",
        "\n",
        "abnormal_data = np.array(normal_data_list).reshape(10485,100)\n",
        "abnormal_data_test = abnormal_data[:2485]\n",
        "\n",
        "from sklearn.model_selection import train_test_split\n",
        "\n",
        "normal_test_tot=np.hstack((normal_data_test, np.zeros(normal_data_test.shape[0]).reshape(-1,1)))\n",
        "abnormal_test_tot=np.hstack((abnormal_data_test, np.ones(abnormal_data_test.shape[0]).reshape(-1,1)))\n",
        "test_tot = np.vstack((normal_test_tot, abnormal_test_tot))\n",
        "\n",
        "x_test, x_vaild, y_test, y_vaild = train_test_split(test_tot[:,:-1], test_tot[:,-1], test_size=0.3)"
      ],
      "metadata": {
        "id": "y1lPh79mgeqg"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "print(x_test.shape)\n",
        "print(x_vaild.shape)\n",
        "print(y_test.shape)\n",
        "print(y_vaild.shape)"
      ],
      "metadata": {
        "id": "ae9mzCVxiQoD",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "36db4a11-7dde-46ae-8237-88b58a5016ae"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "(3479, 100)\n",
            "(1491, 100)\n",
            "(3479,)\n",
            "(1491,)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "# z-score 정규화\n",
        "\n",
        "from sklearn.preprocessing import StandardScaler\n",
        "from sklearn.preprocessing import RobustScaler\n",
        "from sklearn.preprocessing import MinMaxScaler\n",
        "\n",
        "stder = StandardScaler()\n",
        "stder.fit(normal_data_train)\n",
        "x_normal_train_scaled = stder.transform(normal_data_train)\n",
        "x_test_scaled = stder.transform(x_test)\n",
        "x_valid_scaled = stder.transform(x_vaild)"
      ],
      "metadata": {
        "id": "sKSyBc_JiZXc"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# LSTM 입력 데이터 형태로 변경\n",
        "\n",
        "x_normal_train_scaled = x_normal_train_scaled.reshape(x_normal_train_scaled.shape[0], 1, x_normal_train_scaled.shape[1])\n",
        "x_test_scaled = x_test_scaled.reshape(x_test_scaled.shape[0], 1, x_test_scaled.shape[1])\n",
        "x_valid_scaled = x_valid_scaled.reshape(x_valid_scaled.shape[0], 1, x_valid_scaled.shape[1])\n",
        "\n",
        "print(\"훈련 정상 데이터 : \", x_normal_train_scaled.shape)\n",
        "print(\"검증 데이터  : \", x_valid_scaled.shape)\n",
        "print(\"테스트 데이터 : \", x_test_scaled.shape)"
      ],
      "metadata": {
        "id": "L6i-K_VyjojH",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "aa087457-96a5-4faf-9ede-7683dcffbb1a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "훈련 정상 데이터 :  (8000, 1, 100)\n",
            "검증 데이터  :  (1491, 1, 100)\n",
            "테스트 데이터 :  (3479, 1, 100)\n"
          ]
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# LSTM AE 모델 생성"
      ],
      "metadata": {
        "id": "QP-5lGlklans"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# 오토인코더 신경망 구축\n",
        "\n",
        "from keras.models import Sequential\n",
        "from keras.layers import Dense, LSTM, Dropout, RepeatVector, TimeDistributed, Input\n",
        "from keras import regularizers\n",
        "from keras.models import Model\n",
        "\n",
        "def autoencoder_model(X):\n",
        "    inputs = Input(shape=(X.shape[1], X.shape[2]))\n",
        "    L1 = LSTM(32, activation='relu', return_sequences=True, \n",
        "              kernel_regularizer=regularizers.l2(0.00))(inputs)\n",
        "    L2 = LSTM(16, activation='relu', return_sequences=False)(L1)\n",
        "    L3 = RepeatVector(X.shape[1])(L2)\n",
        "    L4 = LSTM(16, activation='relu', return_sequences=True)(L3)\n",
        "    L5 = LSTM(32, activation='relu', return_sequences=True)(L4)\n",
        "    output = TimeDistributed(Dense(X.shape[2]))(L5)    \n",
        "    model = Model(inputs=inputs, outputs=output)\n",
        "    return model"
      ],
      "metadata": {
        "id": "DFltHKr6kfyx"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 모델 생성\n",
        "\n",
        "model = autoencoder_model(x_normal_train_scaled)\n",
        "model.compile(optimizer='adam', loss='mae')\n",
        "model.summary()"
      ],
      "metadata": {
        "id": "YZ8H_irM4NME",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "1995abbb-f51e-4450-daf9-cc7278ea9a93"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Model: \"model_13\"\n",
            "_________________________________________________________________\n",
            " Layer (type)                Output Shape              Param #   \n",
            "=================================================================\n",
            " input_15 (InputLayer)       [(None, 1, 100)]          0         \n",
            "                                                                 \n",
            " lstm_52 (LSTM)              (None, 1, 32)             17024     \n",
            "                                                                 \n",
            " lstm_53 (LSTM)              (None, 16)                3136      \n",
            "                                                                 \n",
            " repeat_vector_14 (RepeatVec  (None, 1, 16)            0         \n",
            " tor)                                                            \n",
            "                                                                 \n",
            " lstm_54 (LSTM)              (None, 1, 16)             2112      \n",
            "                                                                 \n",
            " lstm_55 (LSTM)              (None, 1, 32)             6272      \n",
            "                                                                 \n",
            " time_distributed_14 (TimeDi  (None, 1, 100)           3300      \n",
            " stributed)                                                      \n",
            "                                                                 \n",
            "=================================================================\n",
            "Total params: 31,844\n",
            "Trainable params: 31,844\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "from keras.callbacks import EarlyStopping"
      ],
      "metadata": {
        "id": "UxOR-mno5mPz"
      },
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "source": [
        "# 모델 학습 및 평가\n",
        "\n",
        "nb_epochs = 100\n",
        "batch_size = 64\n",
        "callbacks = [EarlyStopping(monitor='val_loss', patience=50)]\n",
        "history = model.fit(x_normal_train_scaled, x_normal_train_scaled, epochs=nb_epochs, batch_size=batch_size, callbacks=callbacks, validation_split=0.3).history"
      ],
      "metadata": {
        "id": "cjPdC1kAmeKK",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "2112c514-7d80-4bad-b19b-a5a8b8f05a9f"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Epoch 1/100\n",
            "88/88 [==============================] - 7s 17ms/step - loss: 0.0986 - val_loss: 0.0837\n",
            "Epoch 2/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 3/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 4/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 5/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 6/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 7/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 8/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 9/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 10/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0834 - val_loss: 0.0836\n",
            "Epoch 11/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0833 - val_loss: 0.0834\n",
            "Epoch 12/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0829 - val_loss: 0.0832\n",
            "Epoch 13/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0827 - val_loss: 0.0832\n",
            "Epoch 14/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0827 - val_loss: 0.0831\n",
            "Epoch 15/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0826 - val_loss: 0.0831\n",
            "Epoch 16/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0826 - val_loss: 0.0831\n",
            "Epoch 17/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0825 - val_loss: 0.0831\n",
            "Epoch 18/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0825 - val_loss: 0.0830\n",
            "Epoch 19/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0825 - val_loss: 0.0830\n",
            "Epoch 20/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0824 - val_loss: 0.0830\n",
            "Epoch 21/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0824 - val_loss: 0.0830\n",
            "Epoch 22/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0822 - val_loss: 0.0828\n",
            "Epoch 23/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0819 - val_loss: 0.0828\n",
            "Epoch 24/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0818 - val_loss: 0.0827\n",
            "Epoch 25/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0817 - val_loss: 0.0827\n",
            "Epoch 26/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0817 - val_loss: 0.0826\n",
            "Epoch 27/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0817 - val_loss: 0.0826\n",
            "Epoch 28/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0817 - val_loss: 0.0827\n",
            "Epoch 29/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0816 - val_loss: 0.0826\n",
            "Epoch 30/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0816 - val_loss: 0.0826\n",
            "Epoch 31/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0816 - val_loss: 0.0826\n",
            "Epoch 32/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0815 - val_loss: 0.0826\n",
            "Epoch 33/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0815 - val_loss: 0.0826\n",
            "Epoch 34/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0815 - val_loss: 0.0826\n",
            "Epoch 35/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0815 - val_loss: 0.0825\n",
            "Epoch 36/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0814 - val_loss: 0.0825\n",
            "Epoch 37/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0814 - val_loss: 0.0826\n",
            "Epoch 38/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0814 - val_loss: 0.0825\n",
            "Epoch 39/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0814 - val_loss: 0.0825\n",
            "Epoch 40/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0813 - val_loss: 0.0825\n",
            "Epoch 41/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0813 - val_loss: 0.0824\n",
            "Epoch 42/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0813 - val_loss: 0.0824\n",
            "Epoch 43/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0813 - val_loss: 0.0824\n",
            "Epoch 44/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0813 - val_loss: 0.0824\n",
            "Epoch 45/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0812 - val_loss: 0.0824\n",
            "Epoch 46/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0812 - val_loss: 0.0824\n",
            "Epoch 47/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0812 - val_loss: 0.0824\n",
            "Epoch 48/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0812 - val_loss: 0.0824\n",
            "Epoch 49/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0812 - val_loss: 0.0824\n",
            "Epoch 50/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0812 - val_loss: 0.0823\n",
            "Epoch 51/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0811 - val_loss: 0.0823\n",
            "Epoch 52/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0811 - val_loss: 0.0823\n",
            "Epoch 53/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0811 - val_loss: 0.0823\n",
            "Epoch 54/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0811 - val_loss: 0.0823\n",
            "Epoch 55/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0811 - val_loss: 0.0823\n",
            "Epoch 56/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0811 - val_loss: 0.0822\n",
            "Epoch 57/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0811 - val_loss: 0.0822\n",
            "Epoch 58/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0810 - val_loss: 0.0822\n",
            "Epoch 59/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0810 - val_loss: 0.0822\n",
            "Epoch 60/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0810 - val_loss: 0.0821\n",
            "Epoch 61/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0810 - val_loss: 0.0820\n",
            "Epoch 62/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0809 - val_loss: 0.0820\n",
            "Epoch 63/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0809 - val_loss: 0.0819\n",
            "Epoch 64/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0809 - val_loss: 0.0818\n",
            "Epoch 65/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0809 - val_loss: 0.0817\n",
            "Epoch 66/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0808 - val_loss: 0.0817\n",
            "Epoch 67/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0808 - val_loss: 0.0817\n",
            "Epoch 68/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0808 - val_loss: 0.0816\n",
            "Epoch 69/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0808 - val_loss: 0.0816\n",
            "Epoch 70/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 71/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 72/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 73/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 74/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 75/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 76/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0807 - val_loss: 0.0815\n",
            "Epoch 77/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0807 - val_loss: 0.0814\n",
            "Epoch 78/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 79/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 80/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 81/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 82/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 83/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 84/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 85/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 86/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 87/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 88/100\n",
            "88/88 [==============================] - 1s 7ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 89/100\n",
            "88/88 [==============================] - 1s 15ms/step - loss: 0.0806 - val_loss: 0.0814\n",
            "Epoch 90/100\n",
            "88/88 [==============================] - 1s 13ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 91/100\n",
            "88/88 [==============================] - 1s 11ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 92/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0813\n",
            "Epoch 93/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 94/100\n",
            "88/88 [==============================] - 1s 9ms/step - loss: 0.0805 - val_loss: 0.0813\n",
            "Epoch 95/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 96/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 97/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 98/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0813\n",
            "Epoch 99/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n",
            "Epoch 100/100\n",
            "88/88 [==============================] - 1s 8ms/step - loss: 0.0805 - val_loss: 0.0814\n"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "#학습 손실 시각화\n",
        "\n",
        "fig, ax = plt.subplots(figsize=(14, 6), dpi=80)\n",
        "ax.plot(history['loss'], 'b', label='Train', linewidth=2)\n",
        "ax.plot(history['val_loss'], 'r', label='Validation', linewidth=2)\n",
        "ax.set_title('Model loss', fontsize=16)\n",
        "ax.set_ylabel('Loss (mae)')\n",
        "ax.set_xlabel('Epoch')\n",
        "ax.legend(loc='upper right')\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "sOqTsMYAmkma",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 457
        },
        "outputId": "df5e0648-1372-4910-909e-ae1613edc79a"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1120x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "import seaborn as sns\n",
        "sns.set(color_codes=True)\n",
        "\n",
        "X_pred = model.predict(x_normal_train_scaled)\n",
        "X_pred = X_pred.reshape(X_pred.shape[0], X_pred.shape[2]) \n",
        "X_pred = pd.DataFrame(X_pred)\n",
        " \n",
        "Xtrain = x_normal_train_scaled.reshape(x_normal_train_scaled.shape[0], x_normal_train_scaled.shape[2]) \n",
        "scored = pd.DataFrame()\n",
        "scored['Loss_mae'] = np.mean(np.abs(X_pred-Xtrain), axis = 1) \n",
        "\n",
        "Threshold = 0.17\n",
        "\n",
        "plt.figure(figsize=(16,9), dpi=80)\n",
        "plt.title('Loss Distribution', fontsize=16)\n",
        "plt.xlim([0,1])\n",
        "sns.distplot(scored['Loss_mae'], kde= True, color = 'blue');"
      ],
      "metadata": {
        "id": "CfBeCxVDBPOt",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 695
        },
        "outputId": "037ff09d-0cee-45d9-a3ef-e8f49841a993"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/seaborn/distributions.py:2619: FutureWarning: `distplot` is a deprecated function and will be removed in a future version. Please adapt your code to use either `displot` (a figure-level function with similar flexibility) or `histplot` (an axes-level function for histograms).\n",
            "  warnings.warn(msg, FutureWarning)\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 1280x720 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "code",
      "source": [
        "X_pred_3d = model.predict(x_valid_scaled)\n",
        "X_pred = X_pred_3d.reshape(X_pred_3d.shape[0], X_pred_3d.shape[2])\n",
        "X_valid_ = x_valid_scaled.reshape(x_valid_scaled.shape[0], x_valid_scaled.shape[2])\n",
        "\n",
        "mse = np.mean(np.power(X_valid_ - X_pred, 2), axis=1) \n",
        "error_df = pd.DataFrame({'Reconstruction_error' : mse, \n",
        "                         'True_class': y_vaild})\n",
        "\n",
        "precision_rt, recall_rt, threshold_rt = metrics.precision_recall_curve(error_df['True_class'], error_df['Reconstruction_error'])\n",
        "\n",
        "best_cnt_dic = abs(precision_rt-recall_rt)\n",
        "threshold_fixed = threshold_rt[np.argmin(best_cnt_dic)]\n",
        "\n",
        "print('precision:', precision_rt[np.argmin(best_cnt_dic)])\n",
        "print('recall:', recall_rt[np.argmin(best_cnt_dic)])\n",
        "print('threshold:', threshold_fixed)\n",
        "\n",
        "plt.figure(figsize=(8,5))\n",
        "plt.plot(threshold_rt, precision_rt[1:], label='Precision')\n",
        "plt.plot(threshold_rt, recall_rt[1:], label='Recall')\n",
        "plt.xlabel('Threshold'); plt.ylabel('Precision/Recall')\n",
        "plt.legend()\n",
        "plt.show()"
      ],
      "metadata": {
        "id": "iycnNCxPOjch",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 394
        },
        "outputId": "ae6b0eba-7609-44e3-ee7f-cc696c05933e"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "precision: 0.5552731893265566\n",
            "recall: 0.5552731893265566\n",
            "threshold: 0.00026299774085678716\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 576x360 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "# 모델 성능 평가"
      ],
      "metadata": {
        "id": "8MLPN2CSwTjj"
      }
    },
    {
      "cell_type": "code",
      "source": [
        "# AUC/ROC 지표\n",
        "\n",
        "fpr, tpr, Thresholds = roc_curve(y_vaild, mse)\n",
        "roc_auc = auc(fpr, tpr)\n",
        "\n",
        "plt.title('AUC/ROC')\n",
        "plt.plot(fpr, tpr, label='AUC = %0.4f' % roc_auc, linewidth=5 )\n",
        "plt.legend(loc='lower right')\n",
        "plt.plot([0,1], [0,1], 'r--')\n",
        "plt.xlim([-0.001, 1])\n",
        "plt.ylim([0,1.001])\n",
        "plt.ylabel('True Positive Rate')\n",
        "plt.xlabel('False Postive Rate')\n",
        "plt.show();\n"
      ],
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 301
        },
        "id": "OVC5nfd4wSos",
        "outputId": "bda990d6-c362-46b2-f25a-f7a49a6b92d2"
      },
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ],
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZEAAAEcCAYAAAAGD4lRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3dd3hUdfr//+fMpDdSSEJCb4tBWkjoVVoIBIIUXQGXouhiwfZRs6IgYsuu359rQRCXrqvC4gZJEBBBinQFQQkgIdT0RkLaZGbO749sBgIpk0lmMknux3V5XWbmzDl33sa8cs77fe6jUhRFQQghhDCDur4LEEII0XBJiAghhDCbhIgQQgizSYgIIYQwm4SIEEIIs0mICCGEMJuEiBBCCLNJiAghhDCbhIgQwMMPP0yfPn3QarXlXtu0aVO57Y4cOcLQoUONXyuKwvr164mIiKBXr14MHTqUBQsWcO7cuXKfCwsLIzExkaioKLp160ZwcDB9+/Zlzpw5JCQklNs2JSWFF154gX79+tGrVy+mTp3Knj17ym1j6nGFsDQJEdHkXbt2jePHj6NSqfjhhx9q9Nm33nqL9evXs3DhQo4ePcqOHTsYNWoUe/fuNW5z5coVDAYD7du3B+CRRx7hxIkT7Nu3D39/fxYuXGjcNicnh+nTp+Pg4EBsbCyHDx9m9uzZvPDCC2zfvr1GxxXCGuzquwAh6ltMTAw9e/akZ8+exMTEEB4ebtLnLl26xBdffMHXX39Njx49jK9PnDix3HY//vgjw4YNu+vzTk5OhIeH88wzzxhfW7t2LS4uLrz11luo1aV/40VERJCUlER0dDRhYWFcvnzZpOMKYQ1yJiKavC1btjBhwgQmTJjAgQMHyMjIMOlzhw4dokWLFuV+kVdk7969FYZIQUEBsbGxtGnTxvjawYMHGTNmjDFAyoSHh5OUlERiYqLJxxXCGiRERJN2/PhxkpKSCA8Pp1u3brRu3ZrY2FiTPpuTk4Ovr2+V2xQWFvLbb7/Rr18/42urV68mNDSU3r178/PPP/P3v//d+F52dnaF+/Tz8zO+b8pxhbAWCRHRpMXExDBo0CC8vb2B0ktH//3vfwHQaDSUlJSU216n02FnV3oV2NPTk/T09Cr3f+jQIYKDg3FwcDC+NnfuXI4fP87u3btxcnIiMTHR+J6Xl1eF+0xLSzO+b8pxhbAWCRHRZBUVFfHdd99x7NgxBg0axKBBg1i3bh1nz57l7NmzBAQEcP369XKfuXbtGi1btgRgwIABpKSkcPr06UqPsXfv3nKruW4XGBjIwoULeeuttygqKjLu8/vvv8dgMJTb9rvvviMgIID27dubdFwhrEVCRDRZu3btQqPREBcXR0xMDDExMWzbto3Q0FBiYmIYN24c33zzDadOnUJRFBITE1m7di3jxo0DoF27dkyfPp0XXniBI0eOoNVqKS4uJi4ujpUrVwKwf/9+hg8fXmkNgwYNws/Pj6+//hqA2bNnk5eXx8KFC0lPT6e4uJjY2FhWrFjBSy+9hEqlMum4QliLSh5KJZqqRx55hM6dOxMVFVXu9W3btvHWW2+xd+9eYmJiWLNmDcnJyfj4+DBt2jQeffRR48R32f0aGzdu5Nq1a3h4eBASEsKTTz6Joig8//zz5eZYoqKi8Pf357nnnit3vHfeeYcffvgBBwcHkpKSeO+99zhw4ABarZaOHTsyf/58Ro0aZfxMVcft3LmzhUdOiFskRISwkM8++4zs7Gxeeuml+i5FCIuxyuWs6OhoRowYQZcuXTh//nyF2+j1epYsWcKoUaMYPXr0XXcKC9HQtGzZkilTptR3GUJYlFVuNhw5ciR/+ctfmDFjRqXbbN26lStXrrBz505ycnKYNGkSAwYMoFWrVtYoUYg6VzZ3IkRjZpUzkdDQUAICAqrcZtu2bUybNg21Wo23tzejRo0q1+ZBCCGE7bGZ1VnJyckEBgYavw4ICCAlJaUeKxJCCFEdmwkRIYQQDY/NNGAMCAggKSnJ2A/ozjMTU2Vn52MwyIIzHx83MjNv1ncZNkHG4hYZi1ua6lgYCgq4ceU63yToOHc5m86uWqKejzB7fzYTImPHjmXTpk2MGTOGnJwcdu3axRdffFHj/RgMioTI/8g43CJjcYuMxS1NaSwUReHmz8dJ+2IDBcU6Dra6H71aAyrnWu3XKpez3nzzTYYOHUpKSgpz5sxh/PjxAMybN8/YuiEyMpJWrVoxZswYHnjgAZ588klat25tjfKEEKJR0+XkkPzJxySvWIbK04sv/EeUBkgdaHQ3G2Zm3mxSf11UxtfXnfT0vPouwybIWNwiY3FLUxkLXU42lxYtRCkpwWfi/ay/EcCpS9nG9/28nFn16hiz928zl7OEEELUHUNREWonJ+w8vfAKC8c9pA9F7t6c+uhAnR5HVmcJIUQjohgMZO/aycWXXqA4qbQLtc/4CTi0aMGquPg6P56ciQghRCNRnHSd1HVrKEq4gEu3Hqidyk+ax1/OruST5pMQEUKIRiBrWyyZ38agcnKixaOP4d5vACqVqtw2Or3hrs91aeNVq+NKiAghRCOgLyjALbg3vg/NxM7Dw+TPPTiiU62OKyEihBANkEGrJfPbGFzv7YZLUFeaT56KSl3xNHduvrbS+RA3Z/ta1SEhIoQQDUzB+XOkrltNSWoqKnt7XIK63hUgZcERfzm7wstYdUVCRAghGgh9YSEZmzdx48fd2Df3pdULL+ES1LXCbVfFxXP6YmaV+7PT1H6BroSIEEI0EDePHeXG3j14jg6j+aTJqB0dK9wuN19bbYAABLWt3aQ6SIgIIYRN0+flUZx0HZcu9+AxeAhO7dvj2LpNpdvn5mt51oQbCrt38OGR8UG1rk9CRAghbJCiKNw8dpS0Lz8HVLT/+3uo7R0qDZCyOZCqzkDsNGqC2nrxyPggPFwd6qROCREhhLAxJdnZpH2xnvyTJ3Bs154Ws+eitq/4l74p4QGlZx7PPdCzzmuVEBFCCBtSkp3N5UWvoOj1NJ/2IF6jwypcumtqeJSpi0tXFZEQEUIIG2AoKkTt5Iy9lxfe4yJw6x2Kg79/hduaOu9R5p9PD66zy1d3khARQoh6pBgM5Oz6nszYLbR+eSGOLVviHT7+ru3Mue+jbPLcUgECEiJCCFFviq9fI3XdaoouXsS1R0/ULi53bVPTy1ZgnfAoIyEihBD1IDP2WzK3bkHj7EKLeX/FvW8/Y8NEc+82t2Z4lJEQEUKIemAoKsI9tA++f56Onfuthok1ne8oY8l5j6pIiAghhBUYiovJ3PJfXLp1x7XrvXc1TDTnspUl7vuoKQkRIYSwsIKz8aUNE9PTUTs54dr1XmOA2PqcR3UkRIQQwkL0BQVk/GcjN/b9iL2fP63+72Vc7gkya87DFs46KiIhIoQQFnLz+DFu7N+LV1g4PhMnoXZ0NGvOo77mO0whISKEEHVIl5eL9vp1XO4JKm2Y2LEjji1bATWfNLely1aVkRARQog6oCgKeUcPk/blF6jUatpHv8dNLaz6KZP4yxdsfqmuuSREhBCilkqyskj7fB35p37FqUMH/Gc9wk0tNTrrsNU5j+pIiAghRC0YGyYaDLhPmsZX+YHE//sPdPpzJu/Dluc8qiMhIoQQZtAXFpJv0LDq+yu4uXfjjEtrbvzmDGTXaD8NOUBAQkQIIWpE0etJidtG1ratrAsII8PRC5pV/JzzyjTUS1cVkRARQggTFV+9yvXV/0J39TKJrq0p0lT8jPPKNPSzjopIiAghRBXKbgxs9vOPDMg8SZHGkZ0thnLOtS38r2FiVRrTWUdFJESEEKICd7YjGWooId69Pbuah1Kkcarys409OG4nISKEEP9zezsSVUkxQzNP0s61JZdcAtnnHVzlmUdDurejLkmICCEE5e8mb1uQTHjaITx1NynUOHLJJbDKAGmMcx2mkhARQjRpt1+2ctRrGZF5nJ65F8iyd+eLlmFcda74OedN6ZJVVSREhBBNTm6+lmUxh/j1j4xy7Ui65F+me24Chzzv5SfvnujU5X9FNtVLVlWxWogkJiYSFRVFTk4Onp6eREdH065du3LbZGZm8re//Y3k5GR0Oh39+vXj1Vdfxc5Osk4IUTuVtV930RXiq83hsksAp9w7keToS4ajp/F9OeOomtV+Oy9evJjp06cTGRnJli1bWLRoEevXry+3zYoVK+jYsSMrV66kpKSE6dOns3PnTsaNG2etMoUQjUylD31SFO7Nu8iojGPoVWqWt52CXq0pFyBNea7DVFYJkczMTM6cOcOaNWsAiIiIYOnSpWRlZeHt7W3cTqVSkZ+fj8FgQKvVUlJSgr9/xdcjhRCiIqY88Mmj5CZh6UfoWHCda06+fOc3AL1aY3xfLluZziohkpycjL+/PxpN6X8kjUaDn58fycnJ5ULkiSee4Omnn2bw4MEUFhYyY8YMQkJCrFGiEKIRMOV5HW66Ah658i0q4PvmffilWRcUlVouW5nJpiYbtm/fTpcuXVi3bh35+fnMmzeP7du3M3bsWJP34ePjZsEKGxZfX/f6LsFmyFjc0tjGIievmH9+9QunLmRQoqv8mR2Oei3FGgdu2rnwk3dPzrm1pcDZg96dmvPsn3vj6V6zFiailFVCJCAggNTUVPR6PRqNBr1eT1paGgEBAeW2+/zzz3n77bdRq9W4u7szYsQIjhw5UqMQycy8icGg1PW30OD4+rqTnp5X32XYBBmLWxrqWJjzTPIyKsVA35wzDMw6zeetxpLu6EVhyFA+/ksfSoq0AJQUaUn/3783NWq1qlZ/fFslRHx8fAgKCiI2NpbIyEhiY2MJCgoqdykLoFWrVuzbt48ePXqg1Wo5dOgQo0ePtkaJQggbUZvAuJNfcRbj0g7SojiL825taNuhBQsn98HD1QFPd8cmGxx1SaUoilX+bE9ISCAqKorc3Fw8PDyIjo6mQ4cOzJs3jwULFtC9e3euXLnC4sWLycjIQK/X069fPxYuXFijJb5yJlKqof7FaQkyFrfY4ljUZWjcbnDmSQZkn8be3R3/mQ/jHtKn3Pu2OBb1obZnIlYLEWuRECkl/4PcImNxi62NhSkT4eaw06iZpP2dbr52tJw+A43b3b8kbW0s6kuDuJwlhBB3qusAcVbpmVj8O73HD8M3JBhFGYbKhFbtonYkRIQQVlXpzX8mqGwZbv7vv5G6fg26rCzUyR2AYAkQK5EQEUJYnDnzHqbct6G/eZP0jV+Re/AA9i1a0PqlV3Du3LkuSxfVkBARQtSp2k6U16TVyM0TP5N7+CDe4yLwnjARtb3cJGhtEiJCiDpT23kOUwJEdyMHbVISLkFd8Rg0BOdOnXEICDT7mKJ2JESEEHVmVVy8WZ8zpVeVoijkHjxA+tdfobLT0D76PdT2DhIg9UxCRAhhltpetqpJr6qSzAxS16+l4PffcO78J/xnzZFLVzZCQkQIUWPmXrYyp8lhSXY2lxa9CoDf9Jk0Gz4ClVpd42MLyzA5RH766Sfi4uLIyspixYoVnD59mps3bzJgwABL1ieEsBHmnnmY+0wOfUE+GhdX7L28aB55P24hIdj7NK/xfoRlmRTnGzZs4PXXX6ddu3YcO3YMACcnJz744AOLFieEqH+5+Vre3/grz350gNMXM2sUIN07+NQ4QBSdjsy4rVx88QWKr14FwGtMmASIjTLpTGTdunWsXbuWVq1a8dlnnwHQoUMHEhMTLVqcEKL+mXNj4O2XrWqi6PIlUteupvjqFdxC+6Lx8KjR54X1mRQi+fn5xrbtZXeB6nQ67O3tLVeZEKJeWPuyVZmMmG/I2haLxt2dwCefxi1YHkjXEJgUIn369GHlypXMnz/f+Nr69evp16+fxQoTQlhfTSfM6/RpgIqCx8DB+E57EI2ra+32JazGpC6+aWlp/PWvfyUnJ4fU1FRatWqFq6srn376Kb6+vtao02TSxbeUdCi9RcbilurG4v2Nv5p06aounkFuKCokffN/cOvZC9du3VEUxar9ruTnopRVuvj6+fmxefNmTp8+zfXr1wkICKBHjx6oZZmdEI1CTZoidu/gw3MP9KzV8fJ/O0Xq+nXosrOw9/LCtVt3aZjYQJkUIvPnz2f58uX06NGDHj16GF9/6qmn+Pjjjy1WnBDCsmoSHuZOlt9Of/Mm6V9/Se6hn3AICKR11EKcO3Yye3+i/pkUIkeOHKnw9aNHj9ZpMUII66hJeNR2wvx2N0/+Qu7Rw3hHTMR7/ATUsjinwasyRMruAykpKbnrnpCrV68SGCg9a4RoSHLyik2e9wDz7vO4ky4nh+Lr13C9txseAweXNkxsEVCrfQrbUWWIpKSkAKWNz8r+vUxAQABPP/205SoTQtSp3Hwtz7672+TtyybPzaUoCrk/7Sf96y9R2dvfapgoAdKoVBki77zzDgDBwcE88MADVilICFH3arJ0ty5WXmnT00hbv5aC+DM4/6mLNExsxEyaEykLkJs3b5KdnV3uvdatW9d9VUKIWjHnhsG6CA+AkqwsLi9+FZVajd/Mv9Bs6HBpmNiImRQiCQkJvPDCC5w9exaVSlVuPXd8vHnPDxBCWE5NWpXUVXjo8/PRuLpi7+1N88nTcOvdG3tvn1rtU9g+k/48eP311+nXrx9Hjx7Fzc2NY8eO8eCDD/Luu+9auj4hRA2UNUs0NUD++fRgnnugZ60CRNHpyNy6hYsvPU/x1SsAeI0aLQHSRJh0JnL27FlWr16Nvb09iqLg7u7OSy+9REREBJGRkZauUQhRjZos2S1TF0t3iy4lkrJ2NdprV3Hv2w9NM89a7U80PCaFiKOjo7HhopeXF0lJSXh4eJCTk2Pp+oQQJjA1QOzt1NzTpm56XWV88x+yvotD06wZgU89g1uv4FrtTzRMJoVISEgI3333HZMnTyYsLIx58+bh4OBA//79LV2fEKIaufnaagOkrFVJnfaLUqtoNmQozac+gMZFGiY2VSaFyO03Gj7//PN06tSJgoIC7r//fosVJoQwzaq4qhe31PZ+jzL6wkIy/rMRt+BgXLv1wCdysvS7EjV/xrparWbSpElotVo2bdrEjBkzLFGXEMIEVZ2F1NWqK4Cbp06StmE9upxs7Js3x7VbDwkQAZgQIocOHSI+Pp42bdowatQodDod//73v/nss8/w9PSUEBGinlR1A2FddNoF0OXlkv7Vv8k7chiHwJa0nv8Uzh061Hq/ovGoMkRWrlzJ8uXL6dSpExcuXOChhx7i6NGjODg4sHTpUoYPH26lMoUQt6vuDvS6uHwFkP/rSfKOH8N7QiQ+4yegsqvxxQvRyFX5E/H111+zYcMGunXrxsmTJ3nooYd4+eWXmT17tpXKE0KUMfUu9No2TSzJzkabdL20YeKgITh3/hMO/i3M3p9o3KoMkezsbLp16wZAr169cHBwYNasWVYpTAhxS016X5l7FqIoCjf27yVj09eoHBxo/+57qO3tJUBElao9N1UUxfiPo6MjAAbDrb+C5OmGQlhOTW8iNPcGQm1aGqnr11B4Nh7ne4Lw/8scedaHMEmVIVJQUEDXrl2NXyuKYvy6rH+W9M4SwnKsESAlWVlcfv1VVBoNfn+ZTbMhw2TllTBZlSHyww8/1NmBEhMTiYqKIicnB09PT6Kjo2nXrt1d223bto3ly5cbQ2rNmjU0b968zuoQoiEw9Qzk9kfW1jRA9DdvonFzK22YOGUabsEh2Ht716Zs0QRVGSItW7asswMtXryY6dOnExkZyZYtW1i0aBHr168vt83p06f5+OOPWbduHb6+vuTl5eHgIM8gEE2HNR5bq+h0ZMZtJXvHd7SOWohTm7Z4jRxtTrlCmNbFt7YyMzM5c+YMERERAERERHDmzBmysrLKbbd27Vrmzp2Lr68vAO7u7sZ5GCEau7LJc1NamJgbIHnn/+DyG4vJ2rql9MzDS848RO1YZdF3cnIy/v7+aDQaADQaDX5+fiQnJ+N92+lzQkICrVq1YsaMGRQUFDB69Gjmz58v12dFo1aTs4/a3ESYvnkT2Tu+w66ZJ4ELnsWtRy+z9iPE7WzqziG9Xs+5c+dYs2YNWq2WRx99lMDAQCZNmmTyPnx83CxYYcPi6+te3yXYDFsci5y8Yv751S/8fDbNpO1D7vHj2T/3xtPdvLPzQg8XHMeMou2sh7FzcTFrH42NLf5cNDQ1CpHk5GRSU1Pp1atmf8EEBASQmpqKXq9Ho9Gg1+tJS0sjICCg3HaBgYGMHTsWBwcHHBwcGDlyJKdOnapRiGRm3sRgUGpUX2NUp91aGzhbHQtTHx51ew+skiIt6UVak/avLyj4X8PE3rh274HTqHG08fMoHYt82xsPa7PVnwtrU6tVtfrj26QQSUpK4vnnnzc+HvfEiRNs376d/fv389Zbb1X7eR8fH4KCgoiNjSUyMpLY2FiCgoLKXcqC0rmSvXv3EhkZiU6n4/Dhw4SFhZn3nQlhY8x57rm5cx83T54g9fN16G/cwN7PD9fu0jBRWIZJE+uLFi1i+PDh/PLLL9j9r3fOoEGDOHjwoMkHev311/n8888JCwvj888/Z8mSJQDMmzeP06dPAzB+/Hh8fHwYN24ckyZNolOnTkydOrWm35MQNqls3sOUADF38lyXl0vyyuUkffwBGlc32ixchPfYceaWLES1VIqiVHvtp1+/fhw6dAi1Wk3fvn05evQoAKGhoRw/ftziRdaEXM4qJafqt9jKWDz2jx+rDZDatm+/cWA/qRvW4hMxEe/w8Xc1TLSVsbAFMhalrHI5y8fHh8uXL9O+fXvjaxcuXLhrTkMIcbeyy1imBIg5K69KsjLRXr+Oa/ceeAwajPOfuuDg52duuULUiEkhMnfuXP7617/y2GOPodPpiI2N5dNPP2XevHmWrk+IBs2Uxom333VeE4rBwI19P5Lxn42oHB1vNUyUABFWZFKITJ06FU9PT77++msCAgKIiYnhmWeeYdSoUZauT4gGrapH19pp1Kx8cbhZ+9WmppC6bg2F58/hEtQVv7/MloaJol6YFCJ6vZ5Ro0ZJaAhRQ/GXsyt9L6itl1n7LMnK5PKSRag0Gvxnz8Vj0BBZeSXqjUkhMmjQIMaOHcuECRMICQmxdE1CNAq5+dpK50HKJtBrQp+Xh8bdHXtvH3ynPYhbcG/sPM0LIiHqikkhsnr1amJjY3nhhRdQq9WMHz+eiIgIunTpYun6hGhwqmtjUtOlu4aSErLitpK9c7uxYaLnfSPrqlwhasWkEOnatStdu3blpZde4ujRo8TGxjJr1ix8fX3ZunWrpWsUosGobiLdTqOuUYAUJlwgde1qtMlJuA8YiL23T12UKUSdqXHvrA4dOtCxY0cCAwO5dOmSBUoSouExtYliTeZB0jd9TfbO7dh5edHymedx7d6jtmUKUedMCpHc3Fx27NhBbGwsv/76K4MGDeLRRx9l5Eg5pRbCUs8/Vzs60mz4fTSfPA2Ns7O55QlhUSaFyJAhQwgODiYiIoKPPvoIDw8PS9clRINR1TLeMqbcia4vyCd949e49Q7BrUdPvCdEyqorYfNMCpHvv/8eP7mBSYgKVbWMF0ybSL954mdSP9+APi8Xh4AA6NFTAkQ0CJWGyLFjx+jTpw9Q+rCohISECrcbMGCAZSoTogEwZRlvVQGiu3GDtC8/5+bxYzi2bk3Lp5/FqV07C1UrRN2rNESWLFlCbGwsAAsXLqxwG5VKxQ8//GCZyoSwYXW1jDf/t1PknzyBz/1T8A4Lv6thohC2rtKf2LIAAdi9e7dVihGioagqQKpbxluSmUnx9Wu49eiJx8DBuPzpHux9fS1VqhAWZdLzRObPn1/h60899VSdFiNEQ5Cbr61yKW9ly3gVg4Gc3bu4tGghqevWYCgpQaVSSYCIBs2kc+cjR45U+HrZc0WEaEqqW41V0TJebUpyacPEP87j0vVe/KVhomgkqgyRDz74AICSkhLjv5e5evUqgYGBlqtMCBtT3TxIZRPpxoaJ9vb4z3kEj4GDZeWVaDSqDJGUlBQAFEUx/nuZgIAAnn76actVJoSNqS5A7nyglC4vFzt3j9KGiQ8+VNowsZmnNUoVwmqqDJF33nkHgODgYB544AGrFCSErTGlpcntl7AMJVqytn5L9vc7aP23V0sbJg4fYY1ShbC6SkPk2rVrtGrVCii9F+Tq1asVbte6dWvLVCaEjaguQLp38DFewir84w9S1q2iJCUFj4GDpGGiaPQqDZEJEyZw4sQJAEaPHo1KpUJRlHLbqFQq4uOrb/kgRENkyhnI7c8FSd/4Fdnf78DO25uWz76Aa7fu1ipViHpTaYiUBQjA2bNnrVKMELbA1I68d86DqJ2d8Rwxiub3T0Ht5GTpMoWwCWbdHnv16lVUKpXxcpcQjYGp4QGlATJneBtSVv8Lt5BQ3Hr2wjtioqy6Ek2OSTcbPv/88/zyyy8AbN682fhkw02bNlm0OCGspaydu6kB8mhHLZnvLib38EFKUlMBJEBEk2RSiBw6dIhu3boBsHbtWtasWcOmTZv47LPPLFqcENZiSjt3gNCWTkxO3Uvy8mXYeXrR5tXFeI0Js3B1Qtguky5nlZSU4ODgQGpqKjk5OYSEhACQkZFh0eKEsJbq2rmXTaArJ4+QtuEUzadMw2vMWFQajZUqFMI2mRQiQUFBfPrpp1y/fp3hw4cDkJqaipubmyVrE8IqqmvnPqu/Lw45abi5OqAMHIzLPUHY+zS3cpVC2CaTLme99dZbnD9/nuLiYp555hmgdPXWhAkTLFqcEJZW1aNt339yILN90siMXkLq+rW3GiZKgAhhpFLuvPmjgcvMvInB0Ki+JbP4+rqTnp5X32XYhIrGorqVWH66XJ7Q/EZRwgVcunXH/+HZ2Ps0/BsH5efiFhmLUmq1Ch8f868qmbzEd/PmzWzZsoXU1FT8/f2JjIxkypQpZh9YiPpUVYC4l+Qz6+pWtM5OtHhkHu79B8rKKyEqYVKILF++nJiYGObOnUtgYCBJSUn861//Ii0trdJnjQhhyyqaSHfRFVJg50yevSte0x6ied8+2DVrVg/VCdFwmBQimzZtYsOGDbRs2dL42uDBg5k5c6aEiGhw7pxItzPoGJT1K31uxLOhZc78hgoAACAASURBVDh/ezHSpEfbCiFMDJHCwkK8vb3Lvebp6UlRUZFFihLCUu6cSG9VmEp42iF8SnK50rIbLz85SgJEiBowaXXWkCFD+L//+z8uXrxIUVERCQkJREVFMXjwYEvXJ0SduTNARqYfY+b1HWgUAxtbjWHUkv/D07fiR9sKISpm0pnIokWLeOONN5g4cSJ6vR47OzvCw8N59dVXLV2fELWWk1fM+xt/vWsivVDjyLFmQezz6cU9nVrUU3VCNGzVLvHNy8vjypUrtG3bFhcXF7Kzs/Hy8kKtNukkxigxMZGoqChycnLw9PQkOjqadu3aVbjtxYsXuf/++5k+fTovv/xyjY4jS3xLyfLFUreffTjpixiZcZyzbu1IcC3fPPSfTw9uEpex5OfiFhmLUrVd4ltlEvz4448MGTKEKVOmMGzYMI4ePYqPj0+NAwRg8eLFTJ8+nR07djB9+nQWLVpU4XZ6vZ7FixczatSoGh9DiNsZA0RR6HLzEvOufEvXvES8SnLLbddUAkQIS6jyTOT+++9nypQpTJkyhY0bN/Ldd9/x1Vdf1fggmZmZhIWFceTIETQaDXq9nn79+rFz5867JuyXL1+Og4MDBQUFFBQUyJmImZrqX1llNxHGX85GpzfgpitgTPoR/pR/lWRHH77zG0Ca462fuaYWIE3156IiMhalLHqz4dWrV5k5cyYAM2bMYMWKFWYdJDk5GX9/fzT/a1an0Wjw8/MjOTm5XIicPXuWAwcOsH79ej755BOzjlWbwWhsfH3d67sEi8nJK+afX/3CqQsZlOgq7nsF0K4gmfYFSez26c0xz64oqtKz6JB7/Hj2z73xdHe0Vsk2ozH/XNSUjEXtVRkiBsNta+nt7NDr9RYrpKSkhNdee4133nnHGDbmkDORUo39r6yKJsrLNCvJw1ebwwXX1vzm3oErzv7k2t/646Ls7KOkSEt6kdZaJduExv5zURMyFqUseiZSVFTEjBkzjF/n5+eX+xrgiy++qPYgAQEBpKamotfrjZez0tLSCAgIMG6Tnp7OlStXeOyxxwDIzc1FURRu3rzJ0qVLa/RNicavojvOVYqBkBtnGZp5kmK1PYntAtGrNBUGiBCiblQZIm+99Va5r6dOnWrWQXx8fAgKCiI2NpbIyEhiY2MJCgoqdykrMDCQI0eOGL/+6KOPzJoTEU3Dna3bmxfnEJ52kJbFGVxwackO3/7oVaVntPZ2au5p48Uj44MkQISoY1WGyP33319nB3r99deJiorik08+wcPDg+joaADmzZvHggUL6N69e50dSzRuufnlL0G5l+Qz+2osWrU93/oP5oxbe+zsNHRvWxocHdv5yGULISxEWsE3Uo3teu+dq64AXHUF5Nu5ANDjxh889beHsPPwuOuzjW0sakPG4hYZi1IWvU9ECFtR1rpdpzdgZ9AxPONn5l/6Bv+i0sn1M95dKgwQIYRlmfw8ESHqw50Pj2pdmEJ42iG8S/I46dGZHPvSJZpBbaXnlRD1QUJE2LTbA2Rk+jH63Ign286NLwNHc9nl1uq+R8YH1VeJQjRpJoWIVqtl2bJlxMbGkpOTw88//8yBAwe4dOmS8WZEIepCRXMfZQo0jhz17Mo+717o1HbYadQEtZVVV0LUJ5PmRN5++23Onz/Pe++9Z3xMaOfOnfnyyy8tWpxoWsp6XZXNfTjri5iQsp9O+VcBOOTdg93NQ9Gp7ejewYeVLw7nuQd6SoAIUY9MOhPZtWsXO3fuxMXFxdh80d/fn9TUVIsWJ5qWVXHxpf+iKATdvMTo9KM4GkpIcmpebrvuHXzk8pUQNsKkELG3t7+r5UlWVhaenp4WKUo0DRVdunLTFRCWdpjOBddIcmzONr8BZDjemjTv3sGH5x7oWV8lCyHuYNLlrLFjx/Lyyy9z9WrpZYW0tDTeeOMNxo8fb9HiRON2+7LdMu0KkmlXmMwPzUPZ0GqsMUDsNGo5AxHCBpl0JvLcc8/x3nvvMXHiRAoLCwkLC2PatGk8+eSTlq5PNFK5+VrjqivPkjya39Yw8bJLC/LsXAHpdSWErTMpRBwcHHjllVd45ZVXyMrKwsvLyzjBLkRNlU2gqxQDoTlnGZp1gkK1o7FhYlmAdO/gIwEihI0zKUTKLmOVyc/PN/5769at67Yi0eitiouneXE249IOEVicwR8urdjhd6th4u1Ld4UQts2kEBk9ejQqlYrb22yVnYnEx8dbpjLRaF27cJVHrsahVduzxX8I8W7tQKWSSXMhGiCTQuTs2bPlvk5PT+fjjz8mNDTUIkWJxqkkO5tCB1ey1S7s9OvHH66tKdQ4Gd+XMw8hGh6z2p74+vqycOFCwsLCmDBhQl3XJBqB25fvUqJlaOYJQm6cY32rcHDy4ZRH53LbywS6EA2T2b2zLl68SGFhYV3WIhqBOxsmtilIITztIF66m/zi8SeyHe5+prWdRi0BIkQDZVKITJ8+vdxqrMLCQi5cuCBLfIXRneEBMDr9CCE3zpFl784XLcdw1blFhZ+VDrxCNFwmhci0adPKfe3s7Mw999xDu3btLFGTaGDKluzeKU/jwmHPezng3ROduvIfNZkLEaLhqjZE9Ho9hw8fZunSpTg4yCUHUd7tAeKiK2RUxjHOuLfngmtrDntX/shj6cArRONQbYhoNBp++uknublQVGhVXDwoCl1vJjIq/RgOhhKuOfmV26asXYmEhRCNj0mXs2bNmsVHH33E008/jb29vaVrEg1Ebr6WS+euMDX9MJ0KrnPdsTnb/AeS6VDamFPCQ4jGr8oQiY2NJSIigs8//5yMjAzWrFmDt7d3ubOSH3/80dI1ChtUdhnr3sIU2hSmsqt5H35u1gVFVdrTU5bsCtE0VBkiixYtIiIign/84x/Wqkc0AFmJV/l01Q/g1obf3Ttw2SWAm3YuxvclQIRoOqoMkbI2J3379rVKMcK2KXo92Tt3kPrfzYxWO3LRtSV6laZcgEjTRCGalipDxGAwcPjw4XI9s+40YMCAOi9K2J7iq1dIWbua4suXuOjWhh3N+xobJt5OlusK0bRUGSJarZaFCxdWGiIqlYoffvjBIoUJ21GSmcnlN5egdnbh5x7j+T7fGypYrSeXsYRoeqoMEWdnZwmJJqwkKwt7b2/sfXxo9ueHWXKwgKICJ6hgtbcEiBBNk9m9s0TjZSguJuO//yFnz24OhD7A4RzH0kfY3tZx93bS+0qIpsukiXXRdOSf+Z3U9WvQZWRwqXVPjqaDTm2o8jPS+0qIpqvKEDlx4oS16hA2IPWL9dzYsxt7/xZ4L3iRd7elVvuZshsKhRBNk1zOEkb23j64jhrLf5TO/GpCgMg8iBBCXd8FiPqju3GDpBXLuHniZwDsho7mtUt+/Hr5RqWfsdOo6d7BRwJECAHImUiTpCgKeYcOkvbVv1G0xbh0uQf4XzPFKsgz0IUQd5IQaWJKMjNIXb+Wgt9/w6ljJ1rMnotDQCBA6aNsqyBzH0KIO0mINDGF589ReOEPfKfPxHP4CPIKdSzb+Cvxl7NLl/FWQLrxCiEqY7UQSUxMJCoqipycHDw9PYmOjr7ryYjLli1j27ZtqNVq7O3tee655xgyZIi1Smy0tCnJFF+/jntIKO79B+ISdC8F9i788z+nyz3OtiIy9yGEqIrVQmTx4sVMnz6dyMhItmzZwqJFi1i/fn25bXr06MHcuXNxdnbm7NmzzJw5kwMHDuDkVPFNbqJqik5H9s7tZH4bg8bDA7eevVDZ2VFg71Lh42zvJDcRCiGqY5XVWZmZmZw5c4aIiAgAIiIiOHPmDFlZWeW2GzJkCM7OzgB06dIFRVHIycmxRomNzs2LF7ny9lIyvvkPrj170WbhIvKKDby/8VeTAgTkJkIhRPWsciaSnJyMv78/Gk1p11eNRoOfnx/Jycl4e3tX+JmYmBjatGlDixYtrFFio1KSmcEfC/+G2sWFgPlP4h7Sp9yz0Ktz+/PPhRCiKjY5sX706FE++OADVq9eXePP+vi4WaCihqE4PR1HX1/wdcfuyfl49QnB3t0dgGUxh6r9fMg9fjz75954ujtaulSr8/V1r+8SbIaMxS0yFrVnlRAJCAggNTUVvV6PRqNBr9eTlpZGQEDAXdueOHGCF198kU8++YQOHTrU+FiZmTcxGJpWzy9DUSEZ3/yHnL0/0uZvr+LUrj1+I4aTnp4HRXkA/PpHRpX7KJtALynSkl6ktULV1uPr6146FkLG4jYyFqXUalWt/vi2Soj4+PgQFBREbGwskZGRxMbGEhQUdNelrFOnTvHcc8/x4Ycfcu+991qjtAYv/7fTpK5fiy47C88Ro3BocXcw5+ZrZfmuEMIiVIqVWvUmJCQQFRVFbm4uHh4eREdH06FDB+bNm8eCBQvo3r07U6ZM4fr16/j7+xs/9/e//50uXbqYfJymdCaSumEtN/b+iEOLAPxnz8W5U2fje7f/lfX+xl8rXMrbVJbvyl+ct8hY3CJjUapBnIkAdOzYkU2bNt31+meffWb8982bN1urnEbB3tcP7/ET8I6YgNq+4jDIzddWGCCyfFcIURdscmJdVEx3I4e0LzbgMWAgbsEheI8dV+X2Va3IkuW7Qoi6ICHSACiKQu7BA6R//SWKVotL18rni3LztayKi+fslWxKdJU/TEqW7woh6oKEiI0ryUgvbZh45necO/8J/1lzKp08XxUXX20bEyidTJdLWUKIuiAhYuMK//iDwoQE/GY8TLNh96FSl28yUJPwKCNnIUKIuiIhYoOKk5LQJl3HPbQP7v0H4NK1K3bNPMttY054QNNZkSWEsA4JERui6HRkbd9GVuy3aDya4dYrGJWdnTFAyoKjqrbtFbm9jYkEiBCiLkmI2IiiS5dIWbsK7bWruIX2xW/6TFR2pf95zD3rkBsJhRCWJiFiA0oyM7jyzlI0bu4EPvk0bsEh5d6vaYB07+DDS3/pQ0kja18ihLA9EiL1qCQzA3uf5tj7NKfFnEdw7dETjYur8f2anoHcfubh6e7Y6HpgCSFsj4RIPdAXFpLxzSZulDVMbN8BuofyYazMdwghGhYJESvLP32K1A1r0WVn4zJ0BCsPZfHbf67UKDhA5juEELZBQsSKUtev4ca+vTgEBuL97Mu8tDUJyK3xfrp38OG5B3rWfYFCCFFDEiIWpigKKpUKAHv/FnhHTMR7/AQ++O8Zs/ZXdgYihBC2QELEgnQ52aR+vh6PAYNwDwnFOyzc+F785WyT9yPzHkIIWyUhYgGKopC7fx/pm75C0elw7V566cmUmwUlMISt0Ot1ZGeno9M1zlV+aWlqDIaazUU2ZHZ2Dnh5+aLR1O2vfQmROqZNTyN13RoKz8aT6dWK/3r2JeOQAQ7trvaz0pJE2JLs7HScnFxwdW1hvCTbmNjZqdFV0em6MVEUhfz8XLKz02ne/O4GrrUhIVLHss+cI/ePBHb59udXj85g4v988pAoYWt0Om2jDZCmRqVS4erqwc2bOXW+bwmROlB8/Tra5OsoQb3420EtLq0jKbBzrtE+5CFRwhZJgDQelvpvKSFSC4pOR9Z3cWTGfoudpyffBtuBSlWjALl9DkQIIRoaCREzFSVeJGXtarTXr+Hetz++D03nzIpfTPqsTJ4LYZ7c3FwmTQpn4sT7efbZ/zO+vmrVpxQWFvLUU88aX9u8+WvOno1n4cLXAbhy5TLLl3/EhQt/4OHhgYODAw899DBDhw6vk9rWrv0X27ZtBWDcuAnMnv1ohdsNHhxKx46dUKlKnw302mtv0LFjJ7RaLVFRL3DuXOny/7i4H8p97sCBfXzyyQfo9Xq6dAnilVcW4+TkxJ49u1i3brVxu/T0VHr27M3bb/+jTr6v6kiImKEkM4Mr776Fxt2dwKeewdD5Xj6Ki690xZXcXS5E3fj+++3ce283du3awZNPPoO9vb1Jn8vIyOCppx7jiScW8M477wFw40YWBw8erJO6Tp78hT17drFhw9cAPPbYbHr16k2vXr0r3H758tW4uLiUe02tVvPQQzPx9PTk2WefKPdeQUEBf//7Wyxb9hmtW7fh3XeX8uWXG5gzZx733TeK++4bZdx2zpzpjB4dVifflykkRGpAm56Gg69facPEufNw7d6DfMWOZz86UOlnZMWVaOjMfY6NqWpyZh4X9y1PPLGADRvWsn//XkaMGFXl9mW++WYjwcEhjB073viaj09zwsMjalV7mR9++J6wsPE4OjoBEBY2nh9++L7SEKmInZ0dffr0Izk56a73Dh8+yD33BNG6dRsAJk2awptvvs6cOfPKbXfu3FnS09MYPHiY2d9LTUmImEBfUEDGfzZy48A+Wke9inOHDtCtNx/GVt1hV1ZcicbAnGfZ1IROb+D0xUxWxcVX2c7nwoU/yM29QUhIH7KyMomL+9bkEDl//ix9+/Y3adu8vDyefvrxCt9r374Dixe/edfrqakpBN/2CAd//xb8+mvll7effvpx9Ho9/fsPZO7cx3BwqPr3RGpqCv7+t5bm+vu3IC0t9a7t4uK2MHp0uMlnaHVBQqQaN389Sdrn69Dl5OA1Jgytly8rNv5q0v9UsuJKNAYJ12/YxHFiY7cwdux4VCoVw4bdx/vv/4P09DR8ff0qXXlkzookd3d31q79d40/Z6rNm2Px929Bfv5Nli5dxNq1/+Kxx56o/oPV0Gq1fP/9Dj766NM6qNJ0EiJVSFm7itwD+8l18+Hb1uO4luADCcdM+qz0uBKNRceWzSx6JnL7cSpTUlLCrl3bsbd3YPv2OAB0Oh3btm1l1qxH8PT0IiUludxncnJy8PQs/UPuT3+6hzNnfjepDnPORPz9W5Q7fmpqCn5+LSrch79/6euurm5EREzi66+/qLYmf/8WnDhx/I79+5fbZt++PQQGtqRTp87V7q8uSYjcQVEU8gpKWBUXj8u5Iuy8e3LIqxsGlcbkfcg8iGhMHhkfZLU5kcrs37+X1q3bsnz5KuNrv/12ijffXMysWY/Qu3co69atIi0tFT8/f3Jzb7B79/c899xLAEyePI05c2awc+d2xowZC0BWVhY//XTgrnkRc85E7rtvJB988B5TpkwDYMeOOJ599sW7tsvNzcXR0QFHRyd0Oh0//vgDnTv/qdr99+8/gPff/ztXr16hdes2xMRsvutSXlzct4wfP7FGddcFCZHbZF9P5bePVvCTXRvOu7WFZjU7k5BVWKIx8nB1qPdHD8TFfcuYMeHlXuvWrQcGg4ETJ34mODiEBQueJyrqBQwGA4qiMGXKg4SG9gWgeXNfPv54JcuXf8hnny3H2dkJFxcXZsyYVSf19e4dytCh9zFz5oMAjB07zjhHcuDAXg4c2EdU1GtcuXKJf/zjbUCFXq+jW7cePProfON+Hn30L6Snp5KXl8f994+jX78BREW9houLKy+99AovvfQsBoOBzp278Mwzt5Y4p6amcPr0r7zxxrt18v3UhEpRFMXqR7WgzMybGAymf0u5+VpWxZ7B/tQRhqYfR6MY+L55X041M/2U0BbDw9fXnfT0vPouwybIWNxSk7FISblMixZtLVxR/WlKvbPKVPTfVK1W4ePjZvY+m+yZSNmyxatnEwlPP0TbwlQuO7fgO78B5Ni7V/t5uWFQCCGacIiULVsMKs7EvziLbb4DOOXRqdKGiRIaQghxtyYZIsXXrmI4/Qu4tiXerR2XXAIo1DhVuK0tXqoSQghb0aRCxFBSQta2WLLitjJU7Uy8SysMKs1dASJnHUKUuv3xzqJhs9T0d5MJkYzfz3J55We452fym3sHfmgeeteyXTnrEOIWOzsH8vNzcXX1kCBp4MoeSmVnV/e/25pEiJRkpJPxz2gUjTMbA0Zw0bXVXdvYadT1voxRCFvi5eVLdna6RR5kZAvU6qb5eNw632+d79GGZF2+xrojmcRfzqaT32AuurZEq644iaVFiRDlaTR2df4oVVsiS7/rhtVCJDExkaioqP+1IvAkOjqadu3aldtGr9fz5ptvsn//flQqFY899hjTpk2r8bFy0nM4/vG/aHX9dzJahaNzas5Z9/ZVfkZalAghRM1ZLUQWL17M9OnTiYyMZMuWLSxatIj169eX22br1q1cuXKFnTt3kpOTw6RJkxgwYACtWt19+akyy/6/zfSN/55W+iKOenYl3cGz0m1lAl0IIWrHKiGSmZnJmTNnWLNmDQAREREsXbqUrKwsvL29jdtt27aNadOmoVar8fb2ZtSoUWzfvp1HH634CWEVmaQ/T3KLQOJ8Qkh39MK7ku26tGn8j6RVq2UytIyMxS0yFrfIWNR+DKwSIsnJyfj7+6PRlK6G0mg0+Pn5kZycXC5EkpOTCQwMNH4dEBBASkpKjY7V49236AFY77letqs2rQwaGxmLW2QsbpGxqD11fRcghBCi4bJKiAQEBJCamoperwdKJ9DT0tIICAi4a7ukpFuPhkxOTqZFi4p78gshhKh/VgkRHx8fgoKCiI2NBSA2NpagoKByl7IAxo4dy6ZNmzAYDGRlZbFr1y7CwuTClBBC2CqrtYJPSEggKiqK3NxcPDw8iI6OpkOHDsybN48FCxbQvXt39Ho9b7zxBj/99BMA8+bN48EHH7RGeUIIIczQ6J4nIoQQwnpkYl0IIYTZJESEEEKYTUJECCGE2SREhBBCmK3BhUhiYiIPPvggYWFhPPjgg1y6dOmubfR6PUuWLGHUqFGMHj2aTZs2Wb9QKzBlLJYtW8b48eOZMGECkydPZv/+/dYv1ApMGYsyFy9epGfPnkRHR1uvQCsydSy2bdvGhAkTiIiIYMKECWRkZFi3UCswZSwyMzN57LHHmDBhAuHh4bz++uvodDrrF2tB0dHRjBgxgi5dunD+/PkKtzH796bSwDz88MNKTEyMoiiKEhMTozz88MN3bfPf//5XmTt3rqLX65XMzExlyJAhytWrV61dqsWZMhb79u1TCgoKFEVRlPj4eCUkJEQpLCy0ap3WYMpYKIqi6HQ6ZebMmcrzzz+vvPvuu9Ys0WpMGYtTp04p4eHhSlpamqIoipKbm6sUFRVZtU5rMGUs3nzzTePPglarVaZOnarExcVZtU5LO3bsmJKUlKTcd999yrlz5yrcxtzfmw3qTKSskWNERARQ2sjxzJkzZGVllduuskaOjYmpYzFkyBCcnZ0B6NKlC4qikJPTuB4yZOpYAKxcuZLhw4ff9RiCxsLUsVi7di1z587F17f0IUXu7u44OjpavV5LMnUsVCoV+fn5GAwGtFotJSUl+Pv710fJFhMaGnpXh5A7mft7s0GFSFWNHO/crraNHG2dqWNxu5iYGNq0adPoWsmYOhZnz57lwIEDzJ49ux6qtA5TxyIhIYGrV68yY8YM7r//fj755BOLPYO7vpg6Fk888QSJiYkMHjzY+E9ISEh9lFyvzP292aBCRJjv6NGjfPDBB/y///f/6ruUelFSUsJrr73GkiVLjL9UmjK9Xs+5c+dYs2YNGzZsYN++fWzZsqW+y6oX27dvp0uXLhw4cIB9+/Zx/PjxRnflwpIaVIhII8dbTB0LgBMnTvDiiy+ybNkyOnToYO1SLc6UsUhPT+fKlSs89thjjBgxgnXr1rFx40Zee+21+irbIkz9uQgMDGTs2LE4ODjg5ubGyJEjOXXqVH2UbDGmjsXnn3/OxIkTUavVuLu7M2LECI4cOVIfJdcrc39vNqgQkUaOt5g6FqdOneK5557jww8/5N57762PUi3OlLEIDAzkyJEj7N69m927dzNr1iweeOABli5dWl9lW4SpPxcREREcOHAARVEoKSnh8OHD3HPPPfVRssWYOhatWrVi3759AGi1Wg4dOkTnzp2tXm99M/v3Zp0uAbCCCxcuKFOnTlXGjBmjTJ06VUlISFAURVEeffRR5dSpU4qilK7AWbRokTJy5Ehl5MiRyldffVWfJVuMKWMxefJkpV+/fsrEiRON/5w9e7Y+y7YIU8bidh9++GGjXZ1lyljo9Xrl7bffVsaOHauMGzdOefvttxW9Xl+fZVuEKWNx+fJlZfbs2UpERIQSHh6uvP7660pJSUl9ll3nli5dqgwZMkQJCgpSBg4cqIwbN05RlLr5vSkNGIUQQpitQV3OEkIIYVskRIQQQphNQkQIIYTZJESEEEKYTUJECCGE2SRERKPw8MMPN8huzYsWLWLZsmX1XYYQZrOr7wKEuNOIESPIyMgo155k+/btVm+K9/DDD3Py5Ens7OxwcHCgT58+LFq0CD8/P7P2980337Bp0ya+/PJL42tvvPFGXZVbzu1j6OLiwpAhQ3jttddwdXU1q04hKiNnIsImrVixghMnThj/qa+uqosWLeLEiRPs2LGD3Nxc3nnnnXqpwxxlYxgTE8OZM2dYuXJlfZckGiEJEdEg3Lhxg8cff5z+/fvTp08fHn/88Uo7jF6+fJmZM2cSEhJCv379ePbZZ43vJSQkMGfOHPr27UtYWBjbtm0z6fienp6EhYXxxx9/APDLL78wZcoUQkJCmDJlCr/88otx22+++YaRI0cSHBzMiBEj+Pbbb0lISGDx4sWcPHmS4OBgQkNDAYiKiuL9998HIDw8nD179hj3o9Pp6N+/P7///jsAJ0+e5M9//jOhoaFMnDjR5P5Ovr6+DB48mPj4eONrK1euZNSoUQQHBzNu3Di+//574/hUVKdWqyU6Oprhw4czcOBAFi1aRFFRkUnHF42bhIhoEAwGA5MnT2bPnj3s2bMHR0fHSi8FffDBBwwaNIhjx46xb98+Zs6cCUBBQQFz584lIiKCgwcP8v7777NkyRIuXLhQ7fGzsrLYsWMHQUFB5OTk8Pjjj/Pwww9z5MgR5syZw+OPP052djYFBQW8+eabfPbZZ5w4cYKvvvqKoKAgOnbsyJIlS+jVqxcnTpzg+PHjdx1j/Pjxxj5PAAcOHMDLy4t7772X1NRUHn/8cebPn8/Ro0d5+eWXWbBgQYXPR2GZLQAABOZJREFUTLlTSkoK+/fvp02bNsbXWrduzRdffMHPP//MU089xYsvvkhaWlqldb733nskJiYSExPDzp07SUtLk7kcAUiICBv15JNPEhoaSmhoKE888QReXl6EhYXh7OyMm5sb8+fP59ixYxV+1s7OjqSkJNLS0nB0dDT+Nf3jjz/SsmVLpkyZgp2dHV27diUsLKzKtt9vvvkmoaGhREZG4uvry9/+9jd+/PFH2rZty6RJk7CzsyMiIoIOHToYzyLUajV//PEHRUVF+Pn5mdzMb8KECezevZvCwkIAtm7dyvjx4wHYsmULQ4cOZdiwYajVagYNGkS3bt3Yu3dvlWMYHBzMsGHD8Pb2ZsGCBcb3wsPD8ff3R61WM27cONq2bVtpF19FUdi4cSOvvPIKnp6euLm58fjjjxMXF2fS9yUaN5lYFzZp2bJlDBw40Ph1YWEh77zzDvv37+fGjRsA5Ofno9fr73o+yIsvvsgHH3zA1KlTadasGXPmzGHq1Klcv36dU6dOGUMFStuDT5w4sdI6Xn31VaZNm1butbS0tHIP74HSLsGpqam4uLjw/vvvs3r1ahYuXEjv3r15+eWX6dixY7Xfc9u2benYsSN79uzhvvvuY/fu3cTExACQlJTE9u3b77rc1a9fv0r3VzaGR48e5YUXXiA7OxsPDw+g9AFla9as4fr160DpWVp2dnaF+8nKyqKwsJDJkycbX1MUBYPBUO33JBo/CRHRIKxevZrExEQ2btyIr68v8fHxTJo0qcKn8fn6+vLmm28CcPz4cebMmUOfPn0ICAigT58+rFmzpla1+Pn5lXvuApQ+e2HIkCFA6SOJhwwZQlFREf/85z957bXX+Pe//41Kpap23xEREcTGxmIwGOjUqRNt27YFSp/1EBkZafy+aqJv375MnjyZ6OhoPvnkE65fv86rr77K2rVrCQ4ORqPREBkZadz+zjq9vLxwcnIiLi6u0T02VtSeXM4SDUJ+fj6Ojo54eHiQk5PDxx9/XOm23333nXHSvVmzZqhUKtRqNcOHD+fSpUvExMRQUlJCSUkJp06dIiEhoUa1DBs2jEuXLrF161Z0Oh3btm3jwoULDB8+nIyMDHbt2kVBQQEODg64uLigVpf+b+bj40NqaiparbbSfY8bN46ffvqJL7/80vhscICJEyeyZ88e9u/fj16vp7i4mCNHjpj82OdZs2Zx8OBBzp49S2FhISqVyvhcjc2bNxsXDFRUp1qtZtq0abz99ttkZmYCkJqayv79+2s0bqJxkhARDcKsWbMoLi6mf//+PPjgg8a/+ity+vRppk2bRnBwMPPnz2fhwoW0bt0aNzc3Vq1axbZt2xgyZAiDBw/mvffeq/KXekW8vLxYsWIFa9asoV+/fvzrX/9ixYoVeHt7YzAYWLt2LUOGDKFv374cO3aM119/HYD+/fvTqVMnBg8eXOllKD8/P+Ok9rhx44yvBwQE8Mknn/Dpp58yYMAAhg0bxqpVq0y+pOTt7U1kZCTLli2jU6dOzJ07lz//+c8MHDiQ8+fP07t3b+O2FdX54osv0rZtWx544AF69+7N7NmzSUxMrNG4icZJnicihBDCbHImIoQQwmwSIkIIIcwmISKEEMJsEiJCCCHMJiEihBDCbBIiQgghzCYhIoQQwmwSIkIIIcwmISKEEMJs/z/foICVoMxeAgAAAABJRU5ErkJggg==\n"
          },
          "metadata": {}
        }
      ]
    }
  ]
}