GAE-GAT-07-Performance-Evaluation.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e732e07a",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import glob\n",
    "from os import listdir\n",
    "from os.path import isfile, join\n",
    "import os.path as osp\n",
    "import copy\n",
    "import os\n",
    "import pickle\n",
    "from datetime import datetime\n",
    "import random\n",
    "from pm4py.objects.conversion.log import converter as log_converter\n",
    "from pm4py.objects.log.exporter.xes import exporter as xes_exporter\n",
    "from pm4py.objects.log.importer.xes import importer as xes_importer\n",
    "from pm4py.algo.filtering.log.attributes import attributes_filter\n",
    "import numpy as np\n",
    "from sklearn.preprocessing import OneHotEncoder\n",
    "from sklearn import metrics\n",
    "import pandas as pd\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch_geometric.data import Dataset, Data\n",
    "from torch_geometric.loader import DataLoader\n",
    "from torch_geometric.nn import NNConv\n",
    "from torch_geometric.nn import GATv2Conv\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.pyplot import figure"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "acc48670",
   "metadata": {},
   "outputs": [],
   "source": [
    "def nextplot(force=False):\n",
    "    \"\"\"Start a new plot.\n",
    "\n",
    "    In a notebook (or if `force=True`), create a new figure. Otherwise (e.g, in\n",
    "    IPython), clear the current figure.\n",
    "\n",
    "    \"\"\"\n",
    "    inNotebook = \"IPKernelApp\" in get_ipython().config\n",
    "    if force or inNotebook:\n",
    "        plt.figure()  # this creates a new plot\n",
    "    else:\n",
    "        plt.clf()  # and this clears the current one"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "61c5a433",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\exp1\\large-0.1-1.csv'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "77c74930",
   "metadata": {},
   "outputs": [],
   "source": [
    "# import final anomaly scores\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATl1h1nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL1h1wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL1h4nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL1h4wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL2h1nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL2h1wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL2h4nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL2h4wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL3h1nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL3h1wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL3h4nl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp1\\models\\GATL3h4wl_loss.pt'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\exp2\\bestgatmodl_test_loss.pt'\n",
    "\n",
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "id": "cd4ae079",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 248\n",
      "[331, 0.9156626506024096, 0.7657430730478589, 0.8340192043895748, 0.9651424287856072, 0.9892431809450634, 0.9770442041358375, 0.9055317042627061]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, validation_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "c903d8b8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 79,
   "id": "8646340e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[300:400]\n",
    "x_axis = np.arange(300,400)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "id": "c6568ff4",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[310:380]\n",
    "x_axis = np.arange(310,380)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 81,
   "id": "2c21cd66",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[310, 0.9389067524115756, 0.7355163727959698, 0.824858757062147, 0.9609520267757531, 0.9927007299270073, 0.9765684051398338, 0.9007135811009903]\n",
      "false positives 19\n"
     ]
    }
   ],
   "source": [
    "j= 310\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 82,
   "id": "9088f589",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[310, 0.9389067524115756, 0.7355163727959698, 0.824858757062147, 0.9609520267757531, 0.9927007299270073, 0.9765684051398338, 0.9007135811009903]\n"
     ]
    }
   ],
   "source": [
    "j= 310\n",
    "search_more = True\n",
    "while search_more:    \n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if (false_positives/positives > 0.03):\n",
    "        search_more = False\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    best_score = f1_score\n",
    "    precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "    recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "    f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "    macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "    best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                f1_score_on_normal, macro_f1_score]\n",
    "    j += 1   \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "88454c05",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 284\n",
      "[370, 0.9973045822102425, 0.8525345622119815, 0.919254658385093, 0.9756561430201598, 0.9996102883865939, 0.9874879692011549, 0.9533713137931239]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "cfaf3003",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4711097b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[300:400]\n",
    "x_axis = np.arange(300,400)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "787792b9",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZUAAAEGCAYAAACtqQjWAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAApn0lEQVR4nO3deXxV9bnv8c+TnYlMjEkYAjKLqKgQEK3WQutYLSrao7VOrbXUoaf3dtLennPqPR2sxw7X1tk61VqLVSvWAa11qANqAEWZAyiEKUHGAAkZnvvHXmiIISwwO2vv7O/79dqvvfZa67f2dwd2nqzp9zN3R0REpCNkRB1ARES6DhUVERHpMCoqIiLSYVRURESkw6ioiIhIh8mMOkCU+vTp44MHD446hohISpk9e/YGdy9ua1laF5XBgwdTUVERdQwRkZRiZh/sbZkOf4mISIdRURERkQ6joiIiIh1GRUVERDqMioqIiHQYFRUREekwKioiItJh0vo+lQO1fmsdD8z6gPycTPJzMinIiZGfnUlB8Do+L5P8YH5GhkUdWUSkU6ioHIC1W+q4+YVKmkMORZOXHYsXm93POR8XoN0Fafe8YSX5TB5VmtgPICKSICoqB+DIgT1Y9vPT2NnQRG19I9vrm9he3xhMN7Y5b/uuRmpbzFu/tW6P9esamj/a/gvf+xxD+uRH+AlFRA6MisoBMjPysjPJy86Ewk+/vcamZlZt2snkX73Io3Oq+O5JB3/6jYqIdDKdqE8SmbEMhvTJ57jhfXh0zmqawx5bExFJIioqSWbq2DJWb97JrBUfRh1FRGS/qagkmZMP7UtBTiaPzF4ddRQRkf2mopJkumXHOO3wvjz93lq21zdGHUdEZL+oqCShqWPL2LGriZnz10UdRURkv6ioJKHxg3sxsFc3HplTFXUUEZH9oqKShDIyjLOPKuO1ZR+yZvPOqOOIiISmopKkpo4twx0em6sT9iKSOlRUktSg3nmMH9yTR2ZX4a57VkQkNaioJLGpY8tYvmE7c1dtjjqKiEgoKipJ7LQx/cjJzOCR2TphLyKpQUUliRXlZnHyoX154p011Dc2RR1HRGSfVFSS3NRxZWyta+T5hdVRRxER2ScVlSR33PA+lBbl6BCYiKQEFZUkF8swzjxqAC8uqaFmW33UcURE2qWikgKmji2jqdl5/G3dsyIiyU1FJQWMLC3k8AHdeWSOioqIJDcVlRQxdewAFq7dyoI1W6OOIiKyVwktKmZ2ipktNrNKM7umjeVmZjcFy+eZ2dh9tTWzc81svpk1m1l5i/knmtlsM3s3eJ6cyM/W2b505ACyYsaj6mRSRJJYwoqKmcWAm4FTgdHA+WY2utVqpwIjgsflwK0h2r4HnA283GpbG4Az3P1w4GLgjx39maLUKz+bSQeX8Le319DY1Bx1HBGRNiVyT2UCUOnuy919F/AQMKXVOlOA+z1uFtDDzPq119bdF7r74tZv5u5z3X1N8HI+kGtmOYn5aNGYOq6MDbX1vLy0JuooIiJtSmRRGQCsavG6KpgXZp0wbdszFZjr7p+4BtfMLjezCjOrqKlJrV/Okw4uoWdeloYaFpGklciiYm3Ma93d7t7WCdO27Tc1OxT4JfDNtpa7+x3uXu7u5cXFxWE2mTSyMzOYcuQAnluwni07GqKOIyLyCYksKlXAwBavy4A1IdcJ0/YTzKwMeAy4yN2XHUDmpDd1bBm7mpp5Yt4+fxwiIp0ukUXlLWCEmQ0xs2zgPGBGq3VmABcFV4FNBLa4+9qQbfdgZj2AJ4Fr3f3VDv4sSeOwAUWMKCnQUMMikpQSVlTcvRG4CpgJLASmu/t8M5tmZtOC1Z4ClgOVwJ3AFe21BTCzs8ysCjgGeNLMZgbbugoYDvyHmb0dPEoS9fmiYmZMHVfG3JWbWV5TG3UcEZE9WDqPKlheXu4VFRVRx9hv67fWccwvnueKzw3neycfHHUcEUkzZjbb3cvbWqY76lNQaVEux40o5rG5q2luTt8/CkQk+aiopKipYwewevNOZi3/MOooIiIfUVFJUScf2pfCnEz+qhP2IpJEVFRSVG5WjC+O6ccz761je31j1HFERAAVlZQ2dVwZO3Y18fR766KOIiICqKiktPKDejKoV56GGhaRpKGiksLMjLPHDuD15R9StWlH1HFERFRUUt3UsWUA/G2uOpkUkeipqKS4gb3ymDCkF4/MWU0638gqIslBRaULOGdsGSs2bGfOys1RRxGRNKei0gWcenhfcrMy1MmkiERORaULKMzN4pRD+/L3d9ZQ19AUdRwRSWMqKl3E1HFlbK1r5B8L10cdRUTSmIpKF3HssD70Lcrl0Tm6CkxEoqOi0kXEMowzjxrAS0tqqNlWH3UcEUlTKipdyDnjBtDU7Dz+tvZWRCQaKipdyPCSQo4o685f1W2LiERERaWLmTqujEXrtjF/zZaoo4hIGlJR6WLOGNOfrJjxyGwdAhORzqei0sX0zM/m86NKefzt1TQ0NUcdR0TSjIpKFzR1XBkfbt/Fy0tqoo4iImlGRaUL+tzBxfTOz1a3LSLS6VRUuqCsWAZfOrI//1hQzeYdu6KOIyJpREWli5o6toxdTc08MW9t1FFEJI2oqHRRh/Yv4uDSQg01LCKdSkWlizIzpo4bwNurNrOspjbqOCKSJlRUurAzjxxAhqG9FRHpNCoqXVhJUS6fHVnMY3NX09ysoYZFJPESWlTM7BQzW2xmlWZ2TRvLzcxuCpbPM7Ox+2prZuea2Xwzazaz8lbbuzZYf7GZnZzIz5Yqpo4tY+2WOl5f/mHUUUQkDSSsqJhZDLgZOBUYDZxvZqNbrXYqMCJ4XA7cGqLte8DZwMut3m80cB5wKHAKcEuwnbR24uhSCnMz1cmkiHSK/SoqZpZhZkUhV58AVLr7cnffBTwETGm1zhTgfo+bBfQws37ttXX3he6+uI33mwI85O717r4CqAy2k9Zys2JMHVvG3+etYfXmnVHHEZEubp9FxcweNLMiM8sHFgCLzez7IbY9AFjV4nVVMC/MOmHaHsj7paVvfHYo7nDHS8uijiIiXVyYPZXR7r4VOBN4ChgEXBiinbUxr/XZ4r2tE6btgbwfZna5mVWYWUVNTXr0jTWgRzemji3jz2+tonpbXdRxRKQLC1NUsswsi3hRedzdG9j3L3iI7ykMbPG6DFgTcp0wbQ/k/XD3O9y93N3Li4uL97HJruNbnxtGY1Mzf/jXiqijiEgXFqao3A68D+QDL5vZQcDWEO3eAkaY2RAzyyZ+En1Gq3VmABcFV4FNBLa4+9qQbVubAZxnZjlmNoT4yf83Q+RMC4P75HPGEf3546wP2LRd/YGJSGLss6i4+03uPsDdTwtOqH8ATArRrhG4CpgJLASmu/t8M5tmZtOC1Z4ClhM/qX4ncEV7bQHM7CwzqwKOAZ40s5lBm/nAdOLnfZ4BrnT3prA/iHRw5aTh7NjVxD2vam9FRBLD3Ns/kmVm/w7cA2wD7gKOAq5x92cTHy+xysvLvaKiIuoYneqbf6zg9WUf8so1kynKzYo6joikIDOb7e7lbS0Lc/jra8GJ+pOAYuBS4PoOzCed6KpJI9ha18gfX/8g6igi0gWFKSq7r6o6DbjH3d+h7SutJAUcXtadzx1czB9eWcGOXY1RxxGRLiZMUZltZs8SLyozzawQ0ODnKeyqScPZuH0Xf35z1b5XFhHZD2GKyteBa4Dx7r4DyCZ+CExSVPngXkwc2os7Xl5GXYOuZRCRjhPm6q9m4vd8/NjMbgSOdfd5CU8mCXX15BGs31qvPsFEpEOF6ableuDfiV+quwD4tpn9ItHBJLGOHdabowb14LaXltHQpKOZItIxwhz+Og040d3vdve7ifcA/MXExpJEMzOumjScqk07efztfXVWICISTtheinu0mO6egBwSgcmjSjikXxG3vFBJkwbxEpEOEKao/AKYa2b3mtl9wGzg54mNJZ3BzLh68nCWb9jOU++ujTqOiHQBYU7U/xmYCDwaPI4B1M9HF3HKoX0ZVpzPzS9UashhEfnUQh3+cve17j7D3R9393XAwwnOJZ0kI8O4ctJwFq3bxvOLqqOOIyIp7kCHE9Yd9V3Il47oz8Be3fj9P5eyr77gRETac6BFRb95upDMWAbfOmE471Rt4ZXKDVHHEZEUlrm3BWb2BG0XDwN6JyyRRGLquAHc9PxSfvfPSo4fkT6Dl4lIx9prUQFuPMBlkoJyMmN884ShXPfEAt5csZEJQ3pFHUlEUtBei4q7v9SZQSR6540fxM0vVPL7Fyq5f8iEqOOISAo60HMq0gV1y47x9eOG8vKSGt5ZtTnqOCKSglRUZA9fnTiI7t2y+P0LlVFHEZEUpKIieyjMzeLSzwzmuQXrWbRua9RxRCTFHMjVXwC4+5cSkkgid8mxg7nz5eXc/MIyfnf+UVHHEZEU0t6eyo3Ar4h3ybITuDN41ALvJT6aRKVHXjYXHjOYv89bw/Ka2qjjiEgK2WtRcfeXgivAjnL3f3P3J4LHV4DjOi+iROGy44eQk5nBrS8uizqKiKSQMOdUis1s6O4XZjYE0N1xXVyfghzOGz+Ix+auZtXGHVHHEZEUEaao/C/gRTN70cxeBF4AvpPIUJIcvnnCUMzg9pe1tyIi4bR3Rz0A7v6MmY0ARgWzFrl7fWJjSTLo170b54wbyPSKKq6ePILSotyoI4lIkgszRn0e8H3gKnd/BxhkZqcnPJkkhW+dMIymZufOl5dHHUVEUkCYw1/3ALuID84FUAX8NGGJJKkM6p3HlCP686c3VvJhrXZQRaR9YYrKMHe/AWgAcPedaDyVtHLFpGHUNTZx96sa8FNE2hemqOwys24EN0Ka2TBAf7KmkeElhZx2WD/uf+0DtuxsiDqOiCSxMEXlJ8AzwEAz+xPwPPCDMBs3s1PMbLGZVZrZNW0sNzO7KVg+z8zG7qutmfUys+fMbGnw3DOYn2Vm95nZu2a20MyuDZNRwrli0jC21Tdy/2vvRx1FRJLYPouKuz8LnA1cAvwZKHf3F/fVzsxiwM3AqcBo4HwzG91qtVOBEcHjcuDWEG2vAZ539xHEC9zugnMukOPuhwPjgG+a2eB95ZRwDu3fnc+PKuEPr65ge31j1HFEJEmFufrreeBod3/S3f/u7hvM7I4Q254AVLr7cnffBTwETGm1zhTgfo+bBfQws377aDsFuC+Yvg84M5h2IN/MMoFuxC8uUI+IHejKycPZvKOBB99YGXUUEUlSYQ5/DQF+aGb/1WJeeYh2A4BVLV5XBfPCrNNe21J3XwsQPJcE8/8KbAfWAiuBG919Y+tQZna5mVWYWUVNTU2IjyG7jR3Uk88M780d/1pOXUNT1HFEJAmFKSqbgc8DpWb2hJl1D7nttq4Qa93r8d7WCdO2tQlAE9CfeCH8bsvuZT7aiPsd7l7u7uXFxeptZn9dNWkENdvqmV6xat8ri0jaCVNUzN0b3f0K4BHgFT7eO2hPFTCwxesyYE3Iddpruz44REbwXB3M/wrwjLs3uHs18Crh9qhkP0wc2ovyg3py+0vL2dXYHHUcEUkyYYrKbbsn3P1e4ifsnw3R7i1ghJkNMbNs4DxgRqt1ZgAXBVeBTQS2BIe02ms7A7g4mL4YeDyYXglMDraVD0wEFoXIKfvBzLhq8nBWb97J3+aujjqOiCSZvRYVMysKJh8OLuPtZWa9iI+v8r19bdjdG4GrgJnAQmC6u883s2lmNi1Y7SlgOVBJfKyWK9prG7S5HjjRzJYCJwavIX61WAHxsV7eAu5x93khfgayn04YWczhA7pzy4uVNDZpb0VEPmbubZ+qMLO/u/vpZraCT57ncHf/xPmKVFNeXu4VFRVRx0hJz7y3jmkPzOZzBxczqm8RZT27BY88ynp2IzcrFnVEEUkQM5vt7m2eXthrL8XufnrwPCRRwSR1nTS6lK8cPYjXKjfwauUGGpr2/OOkuDBnjyLTcnpADxUdka6qvTHqx+5tGYC7z+n4OJIqMjKMn591OABNzU7NtnqqNu2gatNOVm2MP1dt3sG8qs08895aFR2RNNHeeCq/ameZA5M7OIukqFiG0bd7Ln2751I++JPLm5qd6m118UKzaQdVG3fuUXSefnctjc0fF5287BgPXHY0Ywf17LwPISIdYq/nVNKBzqkkh9ZF5/qnF9Gvezceu+JYzNQhtkiyOaBzKq02cBjxPrg+GvrP3e/vmHiS7mIZRr/u3ejXvRvjB/eiodH5wSPzePLdtZw+pn/U8URkP4Tp++u/gN8Fj0nADcCXEpxL0tjUcWWM6lvIDc8spr5R3cGIpJIwNz+eQ7yblnXufilwBJCT0FSS1mIZxrWnHcLKjTt4YJY6rxRJJWGKyk53bwYagxsiq4GUv0dFktsJI4s5fkQffvfPpWzZoYHBRFJFmKJSYWY9iN/xPhuYA7yZyFAiANeeeghbdjZw84uVUUcRkZDCDNJ1hbtvdvfbiHeLcnFwGEwkoUb3L+Lso8q499X3WbVxR9RxRCSEMHsqmNkYM/sSMBYYbmZnJzaWSNz3Th6JGdz47OKoo4hICGGu/robuBuYCpwRPE5PcC4RAPp178Zlxw/h8bfXMK9qc9RxRGQfwtynMtHdW48tL9Jppp0wjIfeXMXPnlzIQ5dP1A2RIkkszOGv181MRUUiU5ibxb9/YQRvrNjI8wur991ARCITpqjcR7ywLDazeWb2rplpnBLpVOdPGMTQPvn84umFGsNFJImFKSp3AxcCp/Dx+ZQzEhlKpLWsWAY/PHUUy2q285eKVVHHEZG9CFNUVrr7DHdf4e4f7H4kPJlIKyeNLmX84J785rkl1NY3Rh1HRNoQpqgsMrMHzex8Mzt79yPhyURaMTN+dNohbKjdxR0vLYs6joi0IUxR6QbUAyehS4olYkcN6snpY/px579WsH5rXdRxRKSVdi8pNrMYsMHdv99JeUT26Qcnj2Lm/HX8+tkl/PKcMVHHEZEW2t1Tcfcm4nfRiySNQb3zuOiYwUyfvYpF67ZGHUdEWghz+OttM5thZhfqnIoki6snD6cwJ5NfPLUo6igi0kKYotIL+JD4mPQ6pyJJoUdeNldNHs5LS2p4ZemGqOOISGCf3bSoR2JJVhcdM5j7X/+Anz21kL9ffRyxDHXfIhK1MB1KlpnZY2ZWbWbrzewRMyvrjHAi7cnNivH9kw9m4dqtPDZ3ddRxRIRwh7/uAWYA/YEBwBPBPJHInTGmP2PKuvOrZxdT16Dx7EWiFqaoFLv7Pe7eGDzuBYoTnEsklIyM+A2Ra7fU8YdXVkQdRyTthSkqG8zsq2YWCx5fJX7iXiQpTBzamy8cUsqtLy7jw9r6qOOIpLUwReVrwJeBdcBa4JxgnkjSuObUUexsaOL/Pb806igiaS3MGPUr3f1L7l7s7iXufmbYDiXN7JSgy/xKM7umjeVmZjcFy+eZ2dh9tTWzXmb2nJktDZ57tlg2xsxeN7P5QRf9uWFySuobXlLAeeMH8uAbK1leUxt1HJG0tddLis3sP9tp5+7+3+1tOOji5WbgRKAKeMvMZrj7gharnQqMCB5HA7cCR++j7TXA8+5+fVBsrgF+aGaZwAPAhe7+jpn1Bhrayyhdy3e+MJK/zV3NL59ZxO0XlkcdRyQttbensr2NB8DXgR+G2PYEoNLdl7v7LuAhYEqrdaYA93vcLKCHmfXbR9spxAcOI3g+M5g+CZjn7u8AuPuHQTczkiaKC3OYdsIwZs5fz1vvb4w6jkha2mtRcfdf7X4AdxDvrfhS4r/gh4bY9gCg5WhKVcG8MOu017bU3dcGGdcCJcH8kYCb2Uwzm2NmP2grlJldbmYVZlZRU1MT4mNIKrns+KGUFuXwsycX4u5RxxFJO+2eUwnOX/wUmEf8UNlYd/+hu4cZKLyt25tbf8v3tk6Ytq1lAscBFwTPZ5nZ5z+xEfc73L3c3cuLi3VldFfTLTvGd088mLdXbebJd9dGHUck7ey1qJjZ/wBvAduAw939J+6+aT+2XQUMbPG6DFgTcp322q4PDpERPO8ucFXAS+6+wd13AE+hHpbT0tRxZYzqW8gNzyymvlFHQEU6U3t7Kt8lfhf9j4E1ZrY1eGwzszD9jb8FjDCzIWaWDZxH/M78lmYAFwVXgU0EtgSHtNprOwO4OJi+GHg8mJ4JjDGzvOCk/QlAy4sCJE3EMoxrTh3Fyo07eGDWyqjjiKSVvV795e5h7mHZK3dvNLOriP+yjwF3u/t8M5sWLL+N+N7EaUAlsIP4OZu9tg02fT0w3cy+DqwEzg3abDKzXxMvSA485e5PfprPIKnrhJHFHD+iDzc9v5RzxpbRPS8r6kgiacHS+WRmeXm5V1RURB1DEmTBmq188Xf/4hvHD+VHpx0SdRyRLsPMZrt7m9ftf6q9EZFkNrp/EWcfVca9r77Pqo07oo4jkhZUVKRL+97JIzGDG59dHHUUkbSgoiJdWr/u3bjs+CE8/vYa5lVtjjqOSJe3z5EfRVLdtBOG8dCbq/j+w/M47fB+lBblUNo9l9LCXEqLcuiZl02GRo0U6RAqKtLlFeZmcd2UQ/nJjPn85h9LPrE8K2aUBAWmtCiX0qJcSopy6BtMlxblUFKUS2FOJmYqPiLtUVGRtHD6mP6cPqY/9Y1N1GyrZ/3Weqq31rF+ax3rdk9vq2NpdS2vLN3AtvrGT2wjLzsWLziFOR8Vm9KiXApzM8kwiz8yIMMMMyPDINZievdyMyO2e33j4+UZH8/LimUwqm8hmTEdoZbUoqIiaSUnM0ZZzzzKeua1u972+kaqt9WzPig88Uf8dfXWet6p2sy6LXXUNzYnLOtnhvfm9gvLKcjR11RSh/63irQhPyeTITmZDOmTv9d13J0tOxvYvquJ5mbHHZrdg0eL6eb49J7Lg3Wa48/eok2TO8uqa/nF04u44K43uPeS8fTMz+7ETy9y4FRURA6QmdEjL5se7e/0HJBJB5dwUO98rnxwDl++/XX++PWj6dtdY85J8tMBW5EkdeLoUu7/2gTWbqlj6q2vsWLD9n03EomYiopIEps4tDcPXT6RuoYmzr3tNd5bvSXqSCLtUlERSXKHDejO9GnHkJMZ4/w7ZvHmCo1qKclLRUUkBQwrLuDhacdQUpTDhX94g+cXro86kkibVFREUkT/Ht14eNqxHNy3kMv/OJu/zV0ddSSRT1BREUkhvfKzefAbE5kwuBff+cvb3PvqiqgjiexBRUUkxRTkZHLPpeM5aXQpP3liAb/9xxLSeVwkSS4qKiIpKDcrxi0XjOXccWX89h9Lue6JBTQ3q7BI9HTzo0iKyoxlcMM5Y+iRl8Wd/1rB5h27+J9zjyBL/YVJhFRURFKYmfGj0w6hR142/zNzMVvrGrnlgrHkZsWijiZpSn/SiKQ4M+PKScP52VmH8cLiai76w5tsrWuIOpakKRUVkS7igqMP4qbzjmLuqk2cd/ssarbVRx1J0pCKikgXcsYR/bnr4vGs2LCdc297jVUbd0QdSdKMiopIF3PCyGIeuOxoNm7fxbm3vc7S9duijiRpREVFpAsad1BPpk87hmZ3zr39dd5etTnqSJImVFREuqhRfYv467RjKcrN4it3zuKVpRuijiRpQEVFpAsb1DuPv047hkG98vjavW/x9Ltro44kXZzuUxHp4kqKcvnL5cfwtfve4lt/mkOv/GyGFxcwrKSAYcX5DC8pYFhxAQN6dCMjw6KOKylORUUkDXTPy+KPX5/A9LdWsWjdNpbV1PLMe2vZtOPj+1m6ZcUYWpzPsOIChpcUfFRsBvfJIydTN1NKOCoqImkiLzuTSz4zZI95H9bWs6xmO5XVtVRW17KsppbZH2xixjtrPlonw2BQr7x4kQkKze6C071bVmd/DElyCS0qZnYK8P+AGHCXu1/farkFy08DdgCXuPuc9tqaWS/gL8Bg4H3gy+6+qcU2BwELgJ+4+42J/Hwiqa53QQ69C3KYMKTXHvN37Gpkec12ltV8XGwqq2t5aUkNDU0fd1xZXJjD8OICJg7tzRWThqnfMUlcUTGzGHAzcCJQBbxlZjPcfUGL1U4FRgSPo4FbgaP30fYa4Hl3v97Mrgle/7DFNn8DPJ2ozyWSDvKyMzlsQHcOG9B9j/mNTc2s2rRzj0KztLqW3/xjCXNWbuKWC8aSn6MDIOkskf/6E4BKd18OYGYPAVOI70XsNgW43+ODQcwysx5m1o/4Xsje2k4BPhe0vw94kaComNmZwHJgewI/l0jayoxlMKRPPkP65HMipR/N//ObK/nx397j3+54nbsvGU9JYW6EKSVKidxXHQCsavG6KpgXZp322pa6+1qA4LkEwMzyiReX69oLZWaXm1mFmVXU1NTs1wcSkbadP2EQd11UzrLq7Zx182tUVusu/nSVyKLS1rWJrUcR2ts6Ydq2dh3wG3evbW8ld7/D3cvdvby4uHgfmxSRsCaNKuEv35xIfWMTU299nTdXbIw6kkQgkUWlChjY4nUZsCbkOu21XR8cIiN4rg7mHw3cYGbvA98BfmRmV33qTyEioY0p68FjV3yG3gXZfPUPb/DkPN1smW4SWVTeAkaY2RAzywbOA2a0WmcGcJHFTQS2BIe02ms7A7g4mL4YeBzA3Y9398HuPhj4LfBzd/994j6eiLRlYK88Hpl2LGMGdOfKB+dw17+WRx1JOlHCioq7NwJXATOBhcB0d59vZtPMbFqw2lPET6xXAncCV7TXNmhzPXCimS0lfnXYHpcpi0j0euZn88BlR3PqYX356ZMLue6J+TQ17+sItnQFFr/wKj2Vl5d7RUVF1DFEuqzmZuenTy7k7ldXcMqhffnteUdqqOMuwMxmu3t5W8t0p5KIJExGhvGfZ4zmx188hJkL1nHBXW+wafuuqGNJAqmoiEjCXXb8UG7+yljeXb2Fqbe+xsoPNSJlV6WiIiKd4rTD+/Gny47mw+27OPvWV5lXtTnqSJIAKioi0mnGD+7FI986ltysGP92+yz+uWh91JGkg6moiEinGl5SwKNXHMvwkgIuu6+CB99YGXUk6UAqKiLS6UoKc3no8ol8dmQxP3rsXW6cuZh0vhK1K1FREZFI5OdkctdF5Zw3fiC/f6GS705/h12NzVHHkk9JfVSLSGQyYxn84uzDGdCjG796bgnrt9Vx61fHUZSrwb9SlfZURCRSZsbVnx/BjecewRvLN/Ll215n7ZadUceSA6SiIiJJ4ZxxZdxz6XiqNu3k7FteY9G6rVFHkgOgblrUTYtIUlmwZiuX3vsmO+qb+MZnh5KXHSMzw8iMZZAVMzIzMsjKzCArmJcZM7IygudgeXw6g8yM4Hl3u2B+t6wYGRltjbAhYbTXTYvOqYhIUhndv4hHr/gMl91Xwa+fW5KQ9zCDguxMCnMzKczNCp5bTsefi9qa1y3+XJCdqcLUBhUVEUk6A3p046lvH8fOhiYampzGpmYam52GpmYam5zG5uZgvtPQHMxraqahOXgO1mlsCtq0mN/Q1Mz2+ka21jWyra6RbXUNbKtrpKa2nuUbtn80r6Gp/aM4rQtTfk7so72iWEYGMYNYRnxvKRaz+HPG7ueMPV/H9jI/wyjKzWJ4aQHDSwpS4gIGFRURSUpmRl52NL+i3J36xma2BgWnZfHZ/by11bza+kYam5z6hmYam5toanYam52m5ubgOV4Ed89v9nih+3i9+HN7+hblMqK0gBElhYwoLWBkaQHDiwvpnpc8xUZFRUSkFTMjNytGblaMksLOe193p9mhsfnjYrOxdheV1bUsra5l6fptLK2u5c9vrmRnQ9NH7UoKc/YoNiNKChlZWkCPvOzOCx9QURERSRJmFhw2+3jMmaLcLAb3yecLo0s/mtfc7KzevJOl1dtYur6WJetrqazexvSKVezY9XGx6VOQw4iSYI+mtJARJQWMKCmgd0FOwj6DioqISIrJyDAG9spjYK88Jo/as9is2bKTpdW1VK6vZWn1Npasr+WROauprW/8aL3e+dmcddQAfnz66A7PpqIiItJFZGQYZT3zKOuZx6SDSz6a7+6s21rHkvXxQ2iV1bX079EtIRlUVEREujgzo1/3bvTr3o0TRhYn9L10R72IiHQYFRUREekwKioiItJhVFRERKTDqKiIiEiHUVEREZEOo6IiIiIdRkVFREQ6TFoP0mVmNcAHUedopQ+wIeoQ+yGV8qZSVkitvKmUFVIrbzJmPcjd27yLMq2LSjIys4q9jaiWjFIpbyplhdTKm0pZIbXyplJW0OEvERHpQCoqIiLSYVRUks8dUQfYT6mUN5WyQmrlTaWskFp5UymrzqmIiEjH0Z6KiIh0GBUVERHpMCoqncjMcs3sTTN7x8zmm9l1LZZdbWaLg/k3tJg/xsxeD+a/a2a5yZrXzLLM7L4g50IzuzbqrGb2FzN7O3i8b2Zvt2hzrZlVBp/j5M7KeiB5zexEM5sd/Gxnm9nkZM3aot0gM6s1s+91VtYDzRvV9+wA/h9E9h0Lzd316KQHYEBBMJ0FvAFMBCYB/wBygmUlwXMmMA84InjdG4glcd6vAA8F03nA+8DgKLO2WudXwH8G06OBd4AcYAiwLBl+tu3kPQroH0wfBqxO1qwt5j0CPAx8r7OyHuDPNrLv2QFkjew7Fvah4YQ7kcf/J9QGL7OChwPfAq539/pgvepgnZOAee7+TjD/wyTP60C+mWUC3YBdwNaIswJgZgZ8Gdj9F/4U4l/OemCFmVUCE4DXkzGvu89t0Xw+kGtmObv/DZIpazDvTGA5sD3R+Vo7gLyRfc8OIGtk37GwdPirk5lZLNiVrQaec/c3gJHA8Wb2hpm9ZGbjg9VHAm5mM81sjpn9IMnz/pX4L5G1wErgRnffGHHW3Y4H1rv70uD1AGBVi+VVwbxOs595W5oKzO2MgrLb/mQ1s3zgh8B1n9hQJ9nPn22k37P9zBrpdywMFZVO5u5N7n4kUAZMMLPDiO9+9yR+aOn7wPTgL5RM4DjgguD5LDP7fBLnnQA0Af2JH1L6rpkNjTjrbucDf27x2traRALjffLN9i8vAGZ2KPBL4JudEjKwn1mvA37j7rVEZD/zRvo928+skX7HwlBRiYi7bwZeBE4h/lfyox73JtBMvBO5KuAld9/g7juAp4CxSZz3K8Az7t4QHBJ7Fej0PotaZSU4VHA28JcWq1UBA1u8LgPWdE7CPYXMi5mVAY8BF7n7ss5NGRcy69HADWb2PvAd4EdmdlVn5txtP/4vRP49C5k1Kb5j7VFR6URmVmxmPYLpbsAXgEXA3wiOmZrZSCCbeK+kM4ExZpYX/Ac7AViQxHlXApMtLp/4nsyiiLOye9rdq1o0mQGcZ2Y5ZjYEGAG82RlZDyRvsO6TwLXu/mpn5TyQrO5+vLsPdvfBwG+Bn7v775M1LxF+zw4ga2TfsbB0or5z9QPuM7MY8YI+3d3/bmbZwN1m9h7xE28XByfwNpnZr4G3iB+aecrdn0zWvGZ2M3AP8B7xw0v3uPu8KLMGy86j1aEkd59vZtOJ//JoBK5096ZOyrrfeYGrgOHAf5jZfwTzTmpxkUQyZY3a/v5fiPJ7tr8/2yi/Y6GomxYREekwOvwlIiIdRkVFREQ6jIqKiIh0GBUVERHpMCoqIiLSYVRURD4FM4vsrvH9kSo5JfWpqIikkODmPJGkpaIi0sHM7Egzm2Vm88zsMTPrGcz/tpktCOY/FMw7wT4eN2OumRW2sb17zezXZvYC8Eszm2BmrwXrv2ZmBwfrXWJmj5rZM2a21FqMy9NiW30sPm7IFxP8Y5A0pZsfRT4FM6t194JW8+YBV7v7S2b2f4Eid/+Oma0Bhrh7vZn1cPfNZvYE8WEEXjWzAqDO3Rtbbe9e4n2rTXH3JjMrAna4e6OZfQH4lrtPNbNLgP8kPvZKPbAYOM7dVwWHv4YR757mx+7+XAJ/LJLGtKci0oHMrDvQw91fCmbdB3w2mJ4H/MnMvkq8axiIdwj4azP7dtBuj4LSwsMtupHpDjwcdJPzG+DQFus97+5b3L2OeBc0BwXzs4DngR+ooEgiqaiIdJ4vEu+7aRww28wy3f164DLiAy7NMrNRZvaz3YfEWrRtOdjVfwMvuPthwBlAy6FvW46x0sTH/fs1ArOBTh02WdKPiopIB3L3LcQ7Aj0+mHUh8JKZZQAD3f0F4AdAD6DAzIa5+7vu/kugAhjl7v/H3Y8MxthoS3dgdTB9SdhowNeAUWZ2zf5+LpGwdCWJyKeTZ2Ytuyb/NXAxcJuZ5REfUvdSIAY8EBweM+KDWG02s/82s0nE9yoWAE+HeM8biPds+7+Bf4YNGpyPOQ94wsy2uvstYduKhKUT9SIi0mF0+EtERDqMioqIiHQYFRUREekwKioiItJhVFRERKTDqKiIiEiHUVEREZEO8/8BjDaPnkpYrMsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[365:380]\n",
    "x_axis = np.arange(365,380)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "9a956ee5",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[374, 0.9893333333333333, 0.8548387096774194, 0.9171817058096415, 0.976, 0.9984411535463756, 0.9870930456559429, 0.9521373757327922]\n",
      "false positives 4\n"
     ]
    }
   ],
   "source": [
    "j= 374\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d489c20d",
   "metadata": {},
   "source": [
    "GAt Large 15% as the previous cell. Now large 30%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "858e2a70",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\large-thirty\\large-0.3-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\large-thirty\\large30besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "fbe5b09f",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "0f3d6ff4",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 739\n",
      "[785, 0.989821882951654, 0.8790960451977401, 0.9311789347695991, 0.9516711833785004, 0.9962174940898345, 0.9734349734349733, 0.9523069541022862]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "3f48bb8b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "1e962104",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[700:800]\n",
    "x_axis = np.arange(700,800)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "6039c931",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[750:760]\n",
    "x_axis = np.arange(750,760)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "f96a69bb",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[752, 0.99867197875166, 0.8497175141242937, 0.918192918192918, 0.940809968847352, 0.9995271867612293, 0.9692801467216874, 0.9437365324573027]\n",
      "false positives 1\n"
     ]
    }
   ],
   "source": [
    "j= 752\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "b5589cd4",
   "metadata": {},
   "source": [
    "Gat Wide 30%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "ea5562e2",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>{'anomaly': 'Late', 'attr': {'shift_from': 0, ...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id                                              label\n",
       "9000     9001  {'anomaly': 'Late', 'attr': {'shift_from': 0, ..."
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\wide-thirty\\wide-0.3-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\wide-thirty\\wide30besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "id": "e63ac2ef",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "a288466d",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 752\n",
      "[751, 1.0, 0.8025613660618997, 0.8904677323860272, 0.9177046263345195, 1.0, 0.957086522848527, 0.923777127617277]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "0bc157ce",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "id": "098c809e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAEGCAYAAABy53LJAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAtM0lEQVR4nO3dd3xUZdr/8c+VTggplFASOggCCmKkWBAsK7gqltUfay8rNvZx97HuusVn3eL6qLu6q/hg71hXsexaUOyUgEpHQhECCKFKb7l+f8xhzcaQzEAmZ0K+79drXjPnzH0m3zlEr5xz3+c+5u6IiIhEKynsACIiUr+ocIiISExUOEREJCYqHCIiEhMVDhERiUlK2AHqQvPmzb1Dhw5hxxARqVemTp262t1bVF7fIApHhw4dKC4uDjuGiEi9YmZfV7Vep6pERCQmKhwiIhITFQ4REYmJCoeIiMREhUNERGKiwiEiIjFR4RARkZiocFTjk5LV3D+hJOwYIiIJRYWjGh9+VcZdb3/F8vVbw44iIpIwVDiqccHA9rg7D3+8KOwoIiIJQ4WjGoV5mfzo8EIe/WQRUxavDTuOiEhCUOGowW9P7UnzrHTuHT8/7CgiIglBhaMGjdNTuGBAez6av5q533wbdhwRkdCpcETh/AHtaZKewp1vzQs7iohI6FQ4opDXOI3LB3Xi3TmrKFm1Mew4IiKhUuGI0rn925GWnMTv35jDh1+V4e5hRxIRCYUKR5SaZ6Xz0+O68MFXZVz4yGSembwk7EgiIqFQ4YjBT4/vyhe//gG9CrK57fXZnPPAZ7w/b1XYsURE6pQKR4xyMlMZfd7hnNm3kFUbt3Hlk1OZs0KjrUSk4VDh2Adtm2byxzMO4aWrjqRRWjK3/3Nu2JFEROqMCsd+aJaVzpXHduaDr8qYvVxHHSLSMKhw7KcRR7QlLSWJ56aos1xEGoa4Fg4zG2pm88ysxMxuruJ9M7N7g/enm1nfYH2GmU02sy/NbJaZ/U+FbZqa2TtmNj94zovnd6hJbmYapxzamqcnLeGzBWvCjCIiUifiVjjMLBm4DxgG9AB+bGY9KjUbBnQNHiOB0cH67cBx7t4b6AMMNbMBwXs3A+PdvSswPlgO1a2n9aRds0xufOlLtu3cHXYcEZG4iucRRz+gxN0XuvsOYCwwvFKb4cATHjERyDWz1sHypqBNavDwCts8Hrx+HDg9jt8hKtkZqfx+eC+Wrt3KkDsnqLNcRA5o8SwcBcDSCsulwbqo2phZspl9AawC3nH3SUGblu6+AiB4zq/qh5vZSDMrNrPisrKy/f0uNTqyS3MevqiI1Zu28+BHC3XkISIHrHgWDqtiXeV5Ovbaxt13u3sfoBDoZ2a9Yvnh7j7G3YvcvahFixaxbLrPjj+4JX8/ty+7y51ZyzfUyc8UEalr8SwcpUDbCsuFwPJY27j7emACMDRYtdLMWgMEzwl16fZhbXMB+HzJ+lBziIjESzwLxxSgq5l1NLM0YAQwrlKbccCFweiqAcAGd19hZi3MLBfAzBoBJwBzK2xzUfD6IuDVOH6HmOVnZ1CQ24hPSlaHHUVEJC7iVjjcfRcwCngLmAM87+6zzOxKM7syaPYmsBAoAR4Erg7WtwbeN7PpRArQO+7+evDe7cCJZjYfODFYTihn9i1gwldlLFmzJewoIiK1zhrC9OBFRUVeXFxcZz9vxYatDPzTe/xiWHeuOLZznf1cEZHaZGZT3b2o8npdOR4HrXMaUZDbiBnL1EEuIgceFY446VWQzSzNXyUiByAVjjjp1SaHRas3M3HhGsrLD/zTgSLScKhwxMkJPVrSKDWZEWMm0v9P47nr7XlM/Xpd2LFERPabOsfjaOO2nbw/r4xXP1/G+LmrSEkyJv7yeJpnpdd5FhGRWKlzPARNMlI5rXcbHr74CJ69fAC7yp03pq8IO5aIyH5R4agjAzs3o2ebbP723nw+nq+LA0Wk/lLhqEP3jDiM7IxUzn94EsWL14YdR0Rkn6hw1KEu+VmM++nRpKck8bpOWYlIPaXCUcey0lM4qktznpuylJJVm2reQEQkwahwhOCCAe3ZunM35z44ke27dN8OEalfVDhCMKR7Po9efASrNm7n1S8qzzQvIpLYVDhCMrhbCzq1aMw4FQ4RqWdUOEJiZpx4cEsmLVrDxm07w44jIhI1FY4QndCjJTt3Oy9NLQ07iohI1FQ4QlTUPo9jujbnT/+cy00vTmfR6s1hRxIRqZEKR4jMjLvO6c3QXq14rngpN700PexIIiI1UuEIWX6TDO4ZcRjXnXgQkxetZcK8VWFHEhGplgpHgjijbwGN05K55LEpTFui6ddFJHGpcCSIwrxM/vWzQWRnpHLnW/NoCNPdi0j9pMKRQNo2zeT6HxzEpwvW8PDHi8KOIyJSJRWOBHP+gPaccHA+d739FaXrtoQdR0Tke1Q4EoyZcetpPUkyuPyJqWzZsSvsSCIi/0GFIwEV5mXy9/P6MmfFt4yesCDsOCIi/yGuhcPMhprZPDMrMbObq3jfzOze4P3pZtY3WN/WzN43szlmNsvMrq2wza1mtszMvggeJ8fzO4RlSLd8Tuvdhr+9V8J5D03koY8WsmNXedixRERIidcHm1kycB9wIlAKTDGzce4+u0KzYUDX4NEfGB087wKuc/dpZtYEmGpm71TY9i/ufme8sieKP591KL0Ksrnv/QV8UrKGvMw0zjq8MOxYItLAxfOIox9Q4u4L3X0HMBYYXqnNcOAJj5gI5JpZa3df4e7TANx9IzAHKIhj1oTUKC2ZkYM688VvTqQgtxHjvtRMuiISvngWjgJgaYXlUr7/P/8a25hZB+AwYFKF1aOCU1uPmFleVT/czEaaWbGZFZeVle3jV0gMZsYZhxXw4fwynvxsMau+3RZ2JBFpwOJZOKyKdZWvaqu2jZllAS8BP3P3b4PVo4HOQB9gBXBXVT/c3ce4e5G7F7Vo0SLG6Inn6iGd6dM2l1+/Oot+fxzPZY9N0d0DRSQU8SwcpUDbCsuFQOVzLXttY2apRIrG0+7+8p4G7r7S3Xe7eznwIJFTYge8zLQUXrhiIK9ecxSjhnRh/NxVPPSRLhIUkboXz8IxBehqZh3NLA0YAYyr1GYccGEwumoAsMHdV5iZAQ8Dc9z97oobmFnrCotnADPj9xUSS0pyEr3b5nL9Sd0Y2rMVd7/zFe9rUkQRqWNxKxzuvgsYBbxFpHP7eXefZWZXmtmVQbM3gYVACZGjh6uD9UcBFwDHVTHs9g4zm2Fm04EhwM/j9R0S2V3n9KZVdgZPT1wSdhQRaWDiNhwXwN3fJFIcKq57oMJrB66pYruPqbr/A3e/oJZj1kuN01Po36kpH361GncncpAmIhJ/unK8HjusXR6rN22ndN3WsKOISAOiwlGPFbWPjES+8cXpGmElInVGhaMeO7h1NrecfDCfLVzDhHn1+1oVEak/VDjquUuO6kDTxmn8+Z9zWbZep6xEJP5iKhxmlmRm2fEKI7FLSU7ix/3asnD1Zn48ZiLrNu8IO5KIHOBqLBxm9oyZZZtZY2A2MM/Mboh/NInWDSd15+mf9Gf5+q0cd9cE3p71TdiRROQAFs0RR49guo/TiQytbUfkGgtJIEd1ac64UUdTmJfJyCencub9nzBz2YawY4nIASiawpEaTP9xOvCqu+/k+3NOSQLo0SabF68ayKghXZi2ZD2/eHkGu8v1TyUitSuawvF/wGKgMfChmbUHvq12CwlNekoy15/UjdtO78WMZRs46a8faqiuiNSqGguHu9/r7gXufnJw34yviUz1IQnsvH7tuPb4rpSs2sSYD3T3QBGpPdF0jl8bdI6bmT1sZtOA4+ogm+yHpCTj2uO70rF5Y+565yv6/fFd9XmISK2I5lTVpUHn+A+AFsAlwO1xTSW1IinJePySftx1dm8MuOaZafzm1Zl8tXJj2NFEpB6LpnDsmT3vZOBRd/+SvUxAKImnXbNMzjq8kL/8vz6kJifxfPFSfvJ4MVt27Ao7mojUU9EUjqlm9jaRwvGWmTUBdMK8nhncLZ93//tYHrukH0vWbuGml2awYevOsGOJSD0UTeG4DLgZOMLdtwBpRE5XST00oFMzLhrYnte+XM5Jf/mQJWu2hB1JROqZaEZVlRO5peuvzOxO4Eh3nx73ZBI3vz21J3ee3Zs1m7cz6H/f55XPl4UdSUTqkWhGVd0OXEtkupHZwH+Z2Z/iHUziJynJ+NHhhbx45ZF0aJbJ716fzabt6vMQkehEc6rqZOBEd3/E3R8BhgI/jG8sqQu92+Zyx496s3bzDt6dvTLsOCJST0Q7O25uhdc5ccghISlqn0frnAyenbyEud98qylKRKRG0RSOPwGfm9ljZvY4MBX4Y3xjSV1JSjIuOrIDkxatZehfP+LEuz/gmw3bwo4lIgnM3Gv+C9PMWgNHELl+YxLQ3t0nxTlbrSkqKvLi4uKwYyS0+Ss38vnS9fz6lZn0LszlysGdKMjNpE1uBk0yUsOOJyIhMLOp7l5UeX1KNBu7+wpgXIUPm0xkenU5QHRt2YSuLZuwdcdufvf6bC597LtCm52Rwo1Du3P+gPYhJhSRRBHVEcf3NjJb6u5t45AnLnTEEZsNW3eyoGwTy9ZtZfn6rYyfu4rixWs5qGUTjuuez3U/6EZykiYPEDnQ7dcRRxXUg3oAy2mUSt92efRtlwfA+QPac/+EEqaXbuD+CQv4Yul6Rp93ODmZOoUl0hDttXCY2WtUXSAMaBbNh5vZUOAeIBl4yN1vr/S+Be+fDGwBLnb3aWbWFngCaEVkepMx7n5PsE1T4DmgA5H7hJzj7uuiySP7pnF6Cjec1B2AZyYt4Zf/mMHzxUu5fFCnkJOJSBiqO+K4cx/fA8DMkoH7gBOBUmCKmY1z99kVmg0DugaP/sDo4HkXcF1QRJoQmS/rnWDbm4Hx7n67md0cLN9UUx6pHef2b8fjny7m/XmrVDhEGqi9Fg53/2A/P7sfUOLuCwHMbCwwnMjV53sMB57wSEfLRDPLNbPWQWf8iiDHRjObAxQE2w4HBgfbPw5MQIWjTg3u1oKHP17E+DkrOaZrC9JSor0cSEQOBPH8L74AWFphuTRYF1MbM+sAHEZkGDBAy6Cw7BntlV/VDzezkWZWbGbFZWVl+/odpArnD2hPy+wMLnu8mLNGf8r6LTvCjiQidSiehaOqYTeV+0yqbWNmWcBLwM+Cm0lFzd3HuHuRuxe1aNEilk2lBm2bZvLmtcfwu+E9mffNRk79+8c89NFC9mWEnojUP/EsHKVAxSG7hcDyaNuYWSqRovG0u79coc3K4ILEPRcmrqrl3BKFnEapXDiwAw9fXESSGb9/Yw5//tc8yjVlicgBb19GVQHg7qfV8NlTgK5m1hFYBowAzq3UZhwwKuj/6A9scPcVwWirh4E57n53FdtcROT2tRcBr9aQQ+LomK4tmHD9YH75j5k88MECAG4a2o3IP6GIHIiiGVV1JpFhsU8Fyz8mMgy2Wu6+y8xGAW8RGY77iLvPMrMrg/cfAN4kMhS3hMhw3D03iDoKuACYYWZfBOt+6e5vEikYz5vZZcAS4Oyav6bEk5nxxzN64e488MECnvhsMW3zMmnfLJM+7XLp1rIJjdNTyEpPoVVOBs2z0sOOLCL7ocYrx83sQ3cfVNO6RKYrx+vGtp27eWlaKQtWbWbJ2i0sLNvEwtWbv9euUWoyWRkpHNW5GXed00dXoYskqP25cryFmXWqMKy2I6DeZvmejNRkzuv/n/NZrd60ndJ1W9m8fRebtu9iyZotlG3azrJ1W3nli+UsKNvMab3bcPFRHUhN1rBekfogmsLxc2CCmS0MljsAV8QtkRxQmmelV3lqyt1p9UYG789bxR/enMOs5Ru4anAXDmqZpf4RkQQX7bTq6UD3YHGuu2+Pa6paplNVie3Ot+Zx34QS3KFrfhaPXdqPgtxGYccSafD2dqoqmnuOZwI3AKPc/UugnZmdEoeM0kBdf1I3Jv3ieH5/ei+WrtvCtc9+zsxlG3RdiEiCiuZU1aNE7vo3MFguBV4AXo9XKGl48rMzOH9Ae7bs2MUf35zLKX/7mKz0FNo2zaRvu1yuGtyZwrzMsGOKCNGNqip29yIz+9zdDwvWfenuveskYS3Qqar6ZdXGbYyfs4p532xk8ZrNfLZgDanJSVwxqBPXDOlCkkZhidSJ/RlVtcPMGhFcDGhmnYF61cch9Ut+kwx+3O+7G0wuXbuFW8fN4q53vmLj9l2MOq4L2bqdrUhoohn/eCvwL6CtmT0NjAdujGcokYraNs3koYuKOOHgloz5cCGD7nifVz5fpj4QkZBEO6qqGTCAyKSEE919dbyD1Sadqjow7C53pi1Zx+3/nMvUr9dxwsEtefDCwzV8VyRO9mdU1Xigv7u/4e6vu/tqMxsTl5Qi1UhOMo7o0JTnRg7gqsGdeXfOSiYvWht2LJEGJ5pTVR2Bm8zstxXWfa8CidSVlOQkfnpcF3IapTLiwYkMu+cjbnpxOu/OXhl2NJEGIZrCsR44HmhpZq+ZWU58I4nULDMthad/0p+LBnagWeM0xn25nMufLOYPb8zmq5Ubw44nckCLZlSVufsu4Gozuxj4GMiLayqRKPQqyKFXQeTvmG07d/OrV2by4EeLeOzTxbx6zdH0aJMdckKRA1M0RxwP7Hnh7o8BFwNvxymPyD7JSE3mzrN7M+H6weQ0SuXKp6bqyEMkTvY6qsrMst39WzNrWtX77l5veiU1qqphmbZkHZc/XsyazTtIT0licLcW3H/e4Zq+XSRGextVVV3heN3dTzGzRUQu/qv4X527e6f4RK19KhwNz+pN23l20hK+XruFF6eWcvIhrTi+e0u65GfROT+LrPRoztKKNGwxXznu7qcEzx3jGUwkHppnpfPT47tGpm/PzuDhjxfx5oxv/v1+65wMTuzRkv85raeuAxGJUXX3HO9b3YbuPq3244jULjPj+pO6ce0JXfl6zRZKVm1iQdkmPl+ynic++5pDCnI4u6ht2DFF6pXqjtfvquY9B46r5SwicZOanESX/Cy65GcBUF7ujBgzkVtemUmX/CwOa6eBgiLRimrKkfpOfRxSlbWbd3DKvR/RNCuNcdccrVl3RSrZ5ylHgo17mdk5ZnbhnkftRxSpW00bp3HTsO7MXPYtt7wyk0WrN2viRJEo1Di0JJhqZDDQA3gTGEbkIsAn4ppMpA6c1rsN00s38PDHi3h28hIOb5/Hr0/pwSEFORq+K7IX0dzIaQbQG/jc3XubWUvgIXc/tS4C1gadqpKaLCzbxPvzyrjv/RLWbt5B33a5PHfFQFKTozooFzkg7c+pqq3uXg7sMrNsYBVQb67hEIlGpxZZXHZ0R/517TFcelRHpi1Zz8gninX1uUgVoikcxWaWCzxI5N7j04DJ8QwlEpb87Ax+c2oPbjn5YKYsXsdJf/2Q3702W30fIhXUWDjc/Wp3X+/uDwAnAhe5+yXRfLiZDTWzeWZWYmY3V/G+mdm9wfvTK147YmaPmNkqM5tZaZtbzWyZmX0RPE6OJotILC4f1ImPbhzCiCPa8sgni7j8iWJKVm0KO5ZIQohq3gUzOxTosKe9mXVx95dr2CYZuI9IsSkFppjZOHefXaHZMKBr8OgPjA6eAR4D/k7VnfB/cfc7o8kusq/yGqfxh9MPISM1mRenlnLK3z7i+INbcnKv1vRtn0teZhoZqclhxxSpc9GMqnoEOBSYBZQHqx2otnAA/YASd18YfM5YYDhQsXAMB57wyHmAiWaWa2at3X2Fu39oZh1i+jYitSwpyfjtqT0ZOagT946fz3tzV/HG9BUANEpNpqhDHif1bMXZRYWkp6iISMMQzRHHAHfvsQ+fXQAsrbBcyndHE9W1KQBW1PDZo4JrSYqB69x9XeUGZjYSGAnQrl272JKLVNI6pxF/OvNQdpc7H3y1ilXfbmf6sg1MXbyOX70ykzvfnsc5RW254aRuGoklB7xoCsdnZtaj0immaFQ1CL5yD2M0bSobDdwWtLuNyNQol37vQ9zHAGMgMhy3prAi0UhOMo7r3hKAEYC78+mCNTwzaQljPlzIxIVreOySfjRtnBZuUJE4iuZPo8eJFI95QQf2DDObHsV2pUDF2eMKgeX70OY/uPtKd98dDBF+kMgpMZFQmBlHdWnOfef15e/nHsaMZRt4+OOFYccSiatoCscjwAXAUOBU4JTguSZTgK5m1tHM0oj8gTauUptxwIXB6KoBwAZ3r/Y0lZm1rrB4BjBzb21F6tIph7bhxINb8tTEJSxbvzXsOCJxE03hWOLu49x9kbt/vedR00bBfcpHAW8Bc4Dn3X2WmV1pZlcGzd4EFgIlRI4ert6zvZk9C3wGdDOzUjO7LHjrjgpHPUOAn0f5XUXi7vqTulHuzlVPTdW1H3LAimbKkfuBXOA1YPue9TUNx00kmnJE6tLYyUu4+eUZ3HZ6L354SGv1d0i9FfOtYyts+GgVq93dv9chnahUOKQubdu5m2H3fMSi1ZsB6JKfxe9O68mRXZqHnEwkNjHfOjbYKBlY7e43xC2ZyAEmIzWZf157DF8sXc/00vWMnbyUix+dwpOX9aN/p2ZhxxPZb9X2cbj7bqDaW8iKyPdlpCYzoFMzRg7qzD+uPorWuRnc9NJ0pn69Tn0fUu9F0zn+hZmNM7MLzOzMPY+4JxM5QORkpnL7mYeyetMOzhr9KcPu+Yjni5dStnF7zRuLJCD1cYjUkc3bdzHuy+U8/uli5n6zkeyMFMZfN5gWTdLDjiZSpX3uHD8QqHBIIikvd16fsYJrx35OyyYZ3HdeXw5vnxd2LJHv2ecbOZlZoZn9I5jifKWZvWRmhfGJKXLgS0oyTuvdhj+cfgjrt+7gZ899zvNTlvLO7JV8u21n2PFEahTNXFWPAs8AZwfL5wfrToxXKJGG4Nz+7ejYvDGXPDaZG1+KzOLTPCudy47uyGl92lCQ2yjkhCJVi6aP4wt371PTukSmU1WSyLbu2M2azdtZsmYL9743n4kL15KdkcLYkQPp0SY77HjSgO3PPcdXm9n5ZpYcPM4H1tR+RJGGqVFaMoV5mRzZpTljRw7k3f8+lsy0FEY9O41tO3eHHU/ke6IpHJcC5wDfELlPxo+oYhpzEakdXfKzuONHh7KwbDN/e29+2HFEvqfGPg53XwKcVgdZRCQw6KAW/OjwQh74YCH9Ojbj2INahB1J5N/2WjjM7DfVbOfuflsc8ohI4Len9mDmsg1c/OhkDm+Xxx/OOIRurZqEHUuk2lNVm6t4AFwG3BTnXCINXpOMVJ67YiDXDO5C8dfr+PO/5oYdSQSo5ojD3e/a89rMmgDXApcAY4ncrlVE4iynUSrXn9SN5CTjnvHzOef/PuPJy/qRnpIcdjRpwKrtHDezpmb2e2A6kSLT191vcvdVdZJORAC4anBnLj+mI5MXrWXMB7o1rYRrr4XDzP6XyO1fNwKHuPut7r6uzpKJyL9lpCZzyw97cMLB+Tz66WK279IwXQlPdUcc1wFtgF8By83s2+Cx0cy+rZt4IlLR+QPas3bzDsZOXhp2FGnAquvjiOYaDxGpQ4O6tuDoLs259bVZHFKYQ992mhxR6p6Kg0g9kpRk/N8Fh9OscRojnyjmk5LVYUeSBkiFQ6SeaZyewq9P6cHqTTu4/IliVmzYGnYkaWBUOETqoeF9CvjoxiHs2u088vGisONIA6PCIVJPtW2aSf9OTRk/V6PjpW6pcIjUY8d3z2dh2WbGfLgg7CjSgMS1cJjZUDObZ2YlZnZzFe+bmd0bvD/dzPpWeO+R4K6DMytt09TM3jGz+cGzhpVIg3Vq7za0b5bJH9+cy4KyTWHHkQYiboXDzJKB+4BhQA/gx2bWo1KzYUDX4DESGF3hvceAoVV89M3AeHfvCowPlkUapGZZ6YwdOQCAf85YEXIaaSjiecTRDyhx94XuvoPIHFfDK7UZDjzhEROBXDNrDeDuHwJrq/jc4cDjwevHgdPjEV6kvmid04jD2+fx1MQlTP16LTXd1VNkf8WzcBQAFS9vLQ3WxdqmspbuvgIgeM6vqpGZjTSzYjMrLisriym4SH3z21N7sHbLDs4a/RnnPTSJhz5ayKSFa9i8fVfY0eQAVOONnPaDVbGu8p9C0bTZJ+4+BhgDkXuO18ZniiSqQwtzeffnx/L6jOU8+dnX/P6NOQBkpiVz/3l9Gdytyr+vRPZJPI84SoG2FZYLgeX70KaylXtOZwXPGosoArRrlsnVg7vw2S+OZ8otJ/DoxUfQoVljLnlsCo9+oms9pPbEs3BMAbqaWUczSwNGAOMqtRkHXBiMrhoAbNhzGqoa44CLgtcXAa/WZmiRA0GLJukM6Z7P81cO5MjOzbh3/Hy27dSMulI74lY43H0XMAp4C5gDPO/us8zsSjO7Mmj2JrAQKAEeBK7es72ZPQt8BnQzs1Izuyx463bgRDObD5wYLItIFbLSU7h6cBfWbdnJvePnq+NcaoU1hF+koqIiLy4uDjuGSCjcnetfmM5L00oZcURb/njGISQlVdW9KPKfzGyquxdVXh/PznERSQBmxp1nH0qrnHTue38BaSlJ/M9pPTFT8ZB9o8Ih0gCYGdf/oBu7djv/9+FCJi9aS+/CXH59ag+y0vW/AYmNfmNEGggz4+Zh3WmT24g3Z6zgueKlvDitlAGdmnLn2b1pndMo7IhST6iPQ6SB+mzBGt6a9Q1PTfyari2bcPuZh9C7bW7YsSSB7K2PQ7PjijRQAzs349bTevKrHx7MnBXf8qMHPuWd2SvDjiX1gI44RITl67dyxZNTmbFsA8d0bU6/Dk05oUdLurdqok70BkxHHCKyV21yG/HClQP5r+O7snz9Vu5+9yuG3fMRv39jDuXlB/4flxIbHXGIyPcsXbuFe8bP58WppRxSkMMTl/Yjr3Fa2LGkjumIQ0Si1rZpJnecdSh3n9ObeSs3cuboT/mkZHXYsSRBqHCISJWSkowz+xby5KX92LGrnPMemsSTny0OO5YkABUOEalW/07NeO/6Yzm+ez6/fnUWFzw8ieml68OOJSFSH4eIRGXHrnLun1DCY58uxoDjurekZXY6Zxe1pWPzxmHHkzjYWx+HCoeIxKRk1SZ++Y8ZLF27hbKN28lrnMYVgzrRrVUTMlKTObQwh/SU5LBjSi1Q4VDhEKl1X63cyKhnpvHVyk3/XpeZlkxhXiN6tcnh8A55nFPUltRknRWvj1Q4VDhE4uabDdtYum4Lazfv4NOS1Sxdt5UZyzZQtnE7/To25f7z+tI8Kz3smBIjTasuInHTKieDVjkZAJzUs9W/17/y+TJuemk6Vz01leevGKir0A8QOn4Ukbg5/bACfntqT6YsXsc/Pl8WdhypJSocIhJX5xQVUtQ+jxtfnM4zk5bw7badYUeS/aQ+DhGJu2+37eT8hyYxvXQDuZmpnNSjFV3ysyjqkEeftrk6hZWg1MchIqHJzkhl7MgBfPjVal6cupR356zkueKlAPQqyOayoztyyqFtNPqqntARh4iEYu3mHfxz5goe+2Qx81dtonVOBjcP687wPgVhR5OAJjkUkYTStHEa5/Vvz1s/G8QjFxeRn53BtWO/4OGPF4UdTWqgIw4RSQg7d5cz6plpvDVrJe2aZpLXOI3D2uZy3Q8OoklGatjxGiT1cYhIQktNTuKeEYfx9KQlTFuyjvVbdvDUxK95Z/ZKzu3fjrOLCslvkhF2TCHORxxmNhS4B0gGHnL32yu9b8H7JwNbgIvdfVp125rZrcDlQFnwMb909zery6EjDpH66f25q/jbe/OZtmQ9ZnDZUR25cGAH2jXLDDtag1DnRxxmlgzcB5wIlAJTzGycu8+u0GwY0DV49AdGA/2j2PYv7n5nvLKLSGIY0j2fId3z+XTBah75eDEPfbyIF6aW8uzlA+jRJjvseA1WPE9V9QNK3H0hgJmNBYYDFQvHcOAJjxz2TDSzXDNrDXSIYlsRaSCO7NycgZ2a8fnS9fzk8WJOvvcjCvMa0a9jU/KbZNA8K40WTdJpmZ3BER2akpyk60LiKZ6FowBYWmG5lMhRRU1tCqLYdpSZXQgUA9e5+7rKP9zMRgIjAdq1a7ePX0FEEoWZ0bddHu/8fBAvT1vGpEVr+GzBGlZv2s7O3d+dch/asxV/HdGHjFRN7R4v8SwcVZX8yh0qe2tT3bajgduC5duAu4BLv9fYfQwwBiJ9HNFFFpFE1ywrncsHdeLyQZ0AcHe+3bqLsk3beWvWN/zvW/M49W8fc8sPD6ZXQY5m5Y2DeBaOUqBtheVCYHmUbdL2tq27r9yz0sweBF6vvcgiUt+YGTmZqeRkptIlvwuHFORw/QtfcvGjUwDo2y6XIzs3p3lWGs2y0mmelU7PgmyyNcR3n8WzcEwBuppZR2AZMAI4t1KbcUROO40lcipqg7uvMLOyvW1rZq3dfUWw/RnAzDh+BxGpZwYd1IIPbhjC5MVrmblsA699uZz7J5RQXuG8Q1pKEjcP7c5FR3ZQf8g+iFvhcPddZjYKeIvIkNpH3H2WmV0ZvP8A8CaRobglRIbjXlLdtsFH32FmfYicqloMXBGv7yAi9VOjtGSOPagFxx7UgmuGdGF3ubN+yw7WbN7BNxu28dini/nd67MZP3clT17anyQVj5joynERaXDKy51HPlnE79+Yw2VHd+SGk7qpM70KmqtKRCSQlGRcdnRH/l9RWx7+eBFD//oh00vX0xD+kK4NKhwi0iCZGX/+0aE8dVl/Nm7bxWl//4QLHp7MRt1oqkY6VSUiDV7pui08O3kJoycsIKdRKkO653NW30KO6tI87Gih0qkqEZG9KMzL5IaTuvPKNUcxsHMzXp62jBtfnE55+YH/h/W+UOEQEQkcWpjL/ecdzt9+fBjL1m/lppemM+7L5er7qETTqouIVPKDni059qAWvDZ9OS9MLeWZSV/TvVU2jdKSSU0yMtKSGdqzFZ1aZIUdNRTq4xAR2Yvycue+90t4Y8YKlq7dwo7d5f+eFys5yRg1pAujjutywN4rfW99HCocIiIxcHdWb9rBn96cw8ufLyM3M5VOzRtzcOtsjj2oBa1yMsjLTCM5yUhOMswgyYwkM5LNsKQ9y9+tTzKCtol1IaIKhwqHiNSyCfNW8a+Z37Bk7RamLVnHtp3l+/2Z/y4oSUFBCV4nJwWFx4zkpO/WJ1mkOEFkdlgzi8wSa5HlP515KP06Nt2nLLp1rIhILRvcLZ/B3fIB2LhtJ4tWb+abDdtYv3Un5eVOuUO5O+7O7grLkcee9wjeC9YFr3dXeO+7953d5ZE2u93/3XbPn//ukbmYfM86h8bptX9FvAqHiEgtaJKRyqGFuRxaGHaS+Dswe3RERCRuVDhERCQmKhwiIhITFQ4REYmJCoeIiMREhUNERGKiwiEiIjFR4RARkZg0iClHzKwM+HofN28OrK7FOLVFuWKXqNmUKzbKFZv9ydXe3VtUXtkgCsf+MLPiquZqCZtyxS5RsylXbJQrNvHIpVNVIiISExUOERGJiQpHzcaEHWAvlCt2iZpNuWKjXLGp9Vzq4xARkZjoiENERGKiwiEiIjFR4QDMLNnMPjez14Plpmb2jpnND57zKrT9hZmVmNk8MzupjnPdambLzOyL4HFySLkWm9mMIENxsC70fbaXXKHvMzPLNbMXzWyumc0xs4EJsr+qyhXq/jKzbhV+9hdm9q2Z/Szs/VVNrkT4/fq5mc0ys5lm9qyZZcR9f3lwW8OG/AD+G3gGeD1YvgO4OXh9M/Dn4HUP4EsgHegILACS6zDXrcD1VbSr61yLgeaV1oW+z/aSK/R9BjwO/CR4nQbkJsj+qipX6Purws9MBr4B2ifC/tpLrlD3F1AALAIaBcvPAxfHe381+CMOMysEfgg8VGH1cCL/URE8n15h/Vh33+7ui4ASoF8d5tqbOstVQ4ZQ91mM6iSXmWUDg4CHAdx9h7uvJ+T9VU2uvQnj3/F4YIG7f01i/X5VzLU3dZkrBWhkZilAJrCcOO+vBl84gL8CNwLlFda1dPcVAMFzfrC+AFhaoV1psK6ucgGMMrPpZvZIhcPPuswF4MDbZjbVzEYG6xJhn1WVC8LdZ52AMuDR4LTjQ2bWmPD3195yQWL8jgGMAJ4NXoe9v/aWC0LcX+6+DLgTWAKsADa4+9vEeX816MJhZqcAq9x9arSbVLGu1sczV5NrNNAZ6EPkl+SuusxVwVHu3hcYBlxjZoOqaVuX2arKFfY+SwH6AqPd/TBgM5FTB3sTdq6w91fkh5mlAacBL9TUtIp1dZkr1P0VFKrhRE47tQEam9n51W1SG7kadOEAjgJOM7PFwFjgODN7ClhpZq0BgudVQftSoG2F7QuJHBbWSS53X+nuu929HHiQ7w4x6yoXAO6+PHheBfwjyBH2PqsyVwLss1Kg1N0nBcsvEvkfdtj7q8pcCbC/9hgGTHP3lcFy2PurylwJsL9OABa5e5m77wReBo4kzvurQRcOd/+Fuxe6ewcih5/vufv5wDjgoqDZRcCrwetxwAgzSzezjkBXYHJd5drzixA4A5hZl7kAzKyxmTXZ8xr4QZAj1H22t1xh7zN3/wZYambdglXHA7MJ/3esylxh768Kfsx/ng4KdX/tLVcC7K8lwAAzyzQzI/LvOId476/a7uWvrw9gMN+NXmoGjAfmB89NK7S7hchIhHnAsDrO9SQwA5ge/AK0rutcRM6Nfxk8ZgG3JMI+qyZXIuyzPkBxkOEVIC/s/VVNrkTYX5nAGiCnwrpE2F9V5UqE/fU/wFwiRetJIiOm4rq/NOWIiIjEpEGfqhIRkdipcIiISExUOEREJCYqHCIiEhMVDhERiYkKh0gNzGxT2BmiUV9ySv2nwiGSYILJ6kQSlgqHyD4wsz5mNjGY3O4feya3M7P/MrPZwfqxwbpjK9yv4fM9V7hX+rzHzOxuM3sf+LOZ9TOzT4P2n+65wtvMLjazl83sX8G9Fu6o4rOam9lnZvbDOO8GaaB0AaBIDcxsk7tnVVo3Hfipu39gZr8Dst39Z2a2HOjo7tvNLNfd15vZa8Dt7v6JmWUB29x9V6XPewxoDgx3993BtOdb3H2XmZ0AXOXuZ5nZxcBvgMOA7USu/j3a3ZcGp6o6E7mC+Vfu/k4cd4s0YDriEImRmeUAue7+QbDqcSL3toDI1BNPBzOU7ikOnwB3m9l/Bdv9R9Go4AV33x28zgFeMLOZwF+AnhXajXf3De6+jci8V+2D9alEppe4UUVD4kmFQ6R2/RC4DzgcmGpmKe5+O/AToBEw0cy6m9kf9py+qrDt5gqvbwPed/dewKlARoX3tld4vZvIFOkQKVRTgbjeOlhEhUMkRu6+AVhnZscEqy4APjCzJKCtu79P5CZcuUCWmXV29xnu/mcikwp2d/db3L2Pu/fZy4/JAZYFry+ONhpwKdDdzKq754fIftHoDZGaZZpZaYXlu4lMVf2AmWUCC4FLiNyL+qngVJYBfwn6OG4zsyFEjg5mA/+M4mfeATxuZv8NvBdt0KB/ZATwmpl96+73R7utSLTUOS4iIjHRqSoREYmJCoeIiMREhUNERGKiwiEiIjFR4RARkZiocIiISExUOEREJCb/HxgC+7lEi7PMAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[400:800]\n",
    "x_axis = np.arange(400,800)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "cfee6861",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[730:770]\n",
    "x_axis = np.arange(730,770)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "3b2f752f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[745, 1.0, 0.7961579509071505, 0.8865121806298277, 0.9152617568766637, 1.0, 0.95575631225388, 0.9211342464418539]\n",
      "false positives 0\n"
     ]
    }
   ],
   "source": [
    "j= 745\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "64b7ccf7",
   "metadata": {},
   "source": [
    "GAT small 30%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "588c655d",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\small-thirty\\small-0.3-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\small-thirty\\small30besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "1dbe5ca1",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "id": "eb91ae94",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 734\n",
      "[733, 1.0, 0.8407789232531501, 0.9135034225264468, 0.9386584289496911, 1.0, 0.9683587525608923, 0.9409310875436696]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "id": "fc4b3f1c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "97db4046",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[600:800]\n",
    "x_axis = np.arange(600,800)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "id": "8a1bda5e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[690:710]\n",
    "x_axis = np.arange(690,710)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "852b0930",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[701, 1.0, 0.8041237113402062, 0.8914285714285715, 0.9255874673629243, 1.0, 0.9613559322033899, 0.9263922518159806]\n",
      "false positives 0\n"
     ]
    }
   ],
   "source": [
    "j= 701\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "082c3c14",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[759, 0.9697368421052631, 0.8442153493699885, 0.9026331904470299, 0.9392857142857143, 0.9891866478608369, 0.9635905656056789, 0.9331118780263544]\n"
     ]
    }
   ],
   "source": [
    "j= 310\n",
    "search_more = True\n",
    "while search_more:    \n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if (false_positives/positives > 0.03):\n",
    "        search_more = False\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    best_score = f1_score\n",
    "    precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "    recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "    f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "    macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "    best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                f1_score_on_normal, macro_f1_score]\n",
    "    j += 1   \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "09df2e62",
   "metadata": {},
   "source": [
    "GAT wide 15%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "id": "0e726082",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\wide-fifteen\\wide-0.1-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\wide-fifteen\\wide15besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "id": "e645a908",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "04b80987",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 363\n",
      "[391, 0.9821428571428571, 0.8536585365853658, 0.9134045077105575, 0.9746932515337423, 0.9972538250294233, 0.9858444832266822, 0.9496244954686198]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "292717b3",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "c2125d12",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[200:400]\n",
    "x_axis = np.arange(200,400)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "f38477ed",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZgAAAEGCAYAAABYV4NmAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAsiUlEQVR4nO3deXxV9Z3/8dcnO1lICBC2hE0RRGUNiNbdat0qWuvWcddaO9ptOjM6v2k7XaeOo21tp61LW9cqauuC1WoVFTfQhFUEkYAsYUlYEiBAyPb5/XEPGmOWm5Cbc5O8n4/Hfdx7zz3fy/ueh/LhfL/nfL/m7oiIiHS2hLADiIhIz6QCIyIiMaECIyIiMaECIyIiMaECIyIiMZEUdoAwDRgwwEeOHBl2DBGRbmXBggXb3H1gW/v16gIzcuRIiouLw44hItKtmNm6aPZTF5mIiMSECoyIiMSECoyIiMSECoyIiMSECoyIiMSECoyIiMSECoyIiMSECkwH7Kup54ez36dyb03YUURE4pYKTAcs27STR95Zz0V3z2PLzuqw44iIxCUVmA6YNjKX+6+ZxqbKai74/dus3loVdiQRkbijAtNBxx4ygFnXz6C6tp4L75rH0tLKsCOJiMQVFZiDcOSwbP7y9WNJT0nk0nvm8+aqbWFHEhGJGyowB2nUgAz++vVjye+XzjX3F/Hc0s1hRxIRiQsqMJ1gUN80Hv/aMUzIz+amRxfy0PyoJhoVEenRVGA6SXZ6Mg9dezSnjM3j+08v486XV+HuYccSEQmNCkwn6pOSyF2XT+VLU4bxy5c/5L90r4yI9GK9esGxWEhOTOD2L0+kf0YK977xEQ/PX8ekghxOGpvHiYcN5Khh2SQkWNgxRURizmLZjWNmZwB3AonAH9z91iafW/D5WcBe4Cp3X9haWzPLBR4DRgJrgYvcvcLMkoE/AFOIFM4H3f3nreUrLCz0WK5ouXhDJa98UM7cleUs3bgTd+ifkcIJhw3kpLEDOX7MQHIzUmL254uIxIKZLXD3wjb3i1WBMbNE4EPgNKAUKAIudffljfY5C/gGkQJzNHCnux/dWlszuw3Y4e63mtktQD93v9nMvgKc6+6XmFk6sBw4yd3XtpQx1gWmse1V+3l91VbmrtzK66u2sWNPDWZQOKIfv7x4Evn90rskh4jIwYq2wMSyi2w6UOLua4JAs4CZRP7iP2AmkTMNB+abWY6ZDSFydtJS25nASUH7B4DXgJsBBzLMLAnoA9QAu2L4+9qlf2Yq50/O5/zJ+dQ3OO9t3MlrK8v545sfcfV9Rfzl68eS3Sc57JgiIp0mloP8w4ANjd6XBtui2ae1toPcfTNA8JwXbP8LsAfYDKwHbnf3HU1Dmdn1ZlZsZsVbt27tyO86aIkJxqSCHL79+cO4+/KprN2+h689VMz+uvpQ8oiIxEIsC0xzI9lN++Na2ieatk1NB+qBocAo4LtmNvozX+J+j7sXunvhwIED2/jK2Dv2kAHc9uUJzF+zg1v++p4ubRaRHiOWXWSlQEGj9/nApij3SWmlbZmZDXH3zUF3Wnmw/SvAC+5eC5Sb2VtAIbCmM35MLJ0/OZ/SHfu446UPye/Xh++ePjbsSCIiBy2WZzBFwBgzG2VmKcAlwOwm+8wGrrCIGcDOoNurtbazgSuD11cCzwSv1wOnBN+VAcwAPojVj+tsN51yKBcXFvCbV0p4rGh92HFERA5azM5g3L3OzG4CXiRyqfGf3P19M7sh+Pwu4HkiV5CVELlM+erW2gZffSvwuJldS6SoXBhs/y1wH7CMSBfbfe6+NFa/r7OZGT89/0g27dzH/3tqGYOz+3DiYeF34YmIdFRM74OJd115mXK0dlfXcuFd89iwYy9P3HAs44f2DTuSiMinRHuZsqaKiTNZacncf/V0+vZJ5pr7i9i8c1/YkUREOkQFJg4Nzk7jT1dNo2p/HVffV8Su6tqwI4mItJvmIotThw/py+8vm8LV9xVx7f1FnD5+MJlpSWSmJpGZlkRWahJZacmfbEtNIlFznIlIHFGBiWPHjxnI/1wwgf946j2K1la0um//jBQevu5oDh+iMRsRiQ8a5I+zQf7mNDQ4e2rqqNpfR1V1HbuD56r9deyurmV3dR33vrGGpIQEnrnpcwzITA07soj0YPEwF5l0koQEIystmay0ZMhufp9pI3O56O55fP3hBfz5uhmkJGl4TUTCpb+FeoiJBTn874UTKVpbwfee1pQzIhI+ncH0IOdOHMqqst385pUSxg7uy7XHjQo7koj0YjqD6WG+8/nD+MIRg/jZc8t5bWV52w1ERGJEBaaHSUgwfnHRJMYO7ss3HllESXlV2JFEpJdSgemBMlKTuPeKqaQmJ3DdA0VU7q0JO5KI9EIqMD1Ufr907r58Kpsqq7nxkYXU1jeEHUlEehkVmB5s6ohcfnb+kbxVsp2f/m152w1ERDqRriLr4S4sLGBVeRX3vL6GUQMyuOKYkSRoShkR6QIqML3AzWeMY1XZbn747HLueOlDpgzvx9QR/Sgc0Y9Jw3NIT9F/BiLS+fQ3Sy+QmGD8/rKpPLd0M8XrKliwbge/eGnrx5+NH9KXqSMiRefoUbnk9U0LObGI9ASai6wbzEUWCzv31rJwQwUL11VQvLaCxRsq2Vdbjxkce0h/vjw1ny8cMVhnNyLyGdHORaYC00sLTFO19Q18sHk3cz4o468LS9mwYx8ZKYmcPWEIF0zJZ/qoXMw0diMiKjBRUYFpXkODU7R2B39ZUMrz721mT009w3PTuWBKPl+aMoyC3PSwI4pIiFRgoqAC07a9NXW8sGwLf1lQyrw123GHL00Zxi8umhR2NBEJSbQFRvfBSKvSU5L40pR8HvnqDN7495OZOWkoTy7cyNbd+8OOJiJxTgVGopbfL52vTB8OwNLSynDDiEjcU4GRdjlyWDYJBktKd4YdRUTinAqMtEtGahJj8rJ0BiMibVKBkXabkJ/Nkg2VWjVTRFqlAiPtNrEgh4q9tZRW7As7iojEMRUYabeJ+TkALFE3mYi0QgVG2m3s4CxSEhNYsqEy7CgiEsdUYKTdUpISGD+0r64kE5FWqcBIh0zMz2bZxp3UN2igX0SapwIjHTIhP4e9NfWUlFeFHUVE4pQKjHTIxIIcQAP9ItIyFRjpkNEDMshKTdINlyLSIhUY6ZCEBOPIYdks2aCBfhFpngqMdNjEghw+2LKL/XX1YUcRkTikAiMdNjE/m9p6Z8Xm3WFHEZE4pAIjHTbhwEC/brgUkWaowEiHDc1OY0Bmqq4kE5FmqcBIh5kZE/OzWao7+kWkGTEtMGZ2hpmtNLMSM7ulmc/NzH4dfL7UzKa01dbMcs3sJTNbFTz3C7b/k5ktbvRoMLNJsfx9ErnhcvXWKnZX14YdRUTiTMwKjJklAr8FzgTGA5ea2fgmu50JjAke1wO/j6LtLcAcdx8DzAne4+5/dvdJ7j4JuBxY6+6LY/X7JGJiQTbu8N5GncWIyKfF8gxmOlDi7mvcvQaYBcxsss9M4EGPmA/kmNmQNtrOBB4IXj8AnNfMn30p8Gin/hpp1oRg6n51k4lIU+0qMGaWYGZ9o9x9GLCh0fvSYFs0+7TWdpC7bwYInvOa+bMvpoUCY2bXm1mxmRVv3bo1yp8iLcnNSKEgt4/u6BeRz2izwJjZI2bW18wygOXASjP7tyi+25rZ1nTq3Zb2iaZt83+o2dHAXndf1tzn7n6Puxe6e+HAgQOj+Uppw8T8HN3RLyKfEc0ZzHh330WkK+p5YDiRMY62lAIFjd7nA5ui3Ke1tmVBNxrBc3mT77wEdY91qYn5OWys3Me2qv1hRxGROBJNgUk2s2QiBeYZd68lurOJImCMmY0ysxQif/HPbrLPbOCK4GqyGcDOoNurtbazgSuD11cCzxz4MjNLAC4kMmYjXWRCfjaAuslE5FOiKTB3A2uBDOB1MxsB7GqrkbvXATcBLwIrgMfd/X0zu8HMbgh2ex5YA5QA9wL/3FrboM2twGlmtgo4LXh/wAlAqbuvieJ3SSc5clg2CQaL1U0mIo2Ye/tXJDSzpKAIdGuFhYVeXFwcdowe4Qu/fJ0hOWncf/X0sKOISIyZ2QJ3L2xrv2gG+b8VDPKbmf3RzBYCp3RKSukxJgR39HfkHywi0jNF00V2TTDIfzowELiaT3dLiTCxIIcde2oordgXdhQRiRPRFJgDlwyfBdzn7kto/jJi6cUmBjdcauJLETkgmgKzwMz+QaTAvGhmWUBDbGNJdzN2cBYpiQm6o19EPpYUxT7XApOANe6+18z6E+kmE/lYSlIC44f2ZbHWhhGRQJsFxt0bzCwf+IqZAcx192djnky6nYn52TyxoJT6BicxQb2oIr1dNFeR3Qp8i8g0McuBb5rZz2MdTLqfCfk57K2pZ/XWqrCjiEgciKaL7Cxgkrs3AJjZA8Ai4D9iGUy6n4nBEsqLN1Ry2KCscMOISOiinU05p9Hr7BjkkB5g9IAMslKTNGWMiADRncH8HFhkZq8SuTz5BHT2Is1ISDCOHKYllEUkos0zGHd/FJgBPBk8jgE+inEu6aYmFuSwYvMu9tfVhx1FREIWzRnMgYW9Pp4J2czeJTJtv8inTMzPprbeWbF5N5OCMRkR6Z06umSyrkGVZk0IiorGYUSkowVGMxpKs4ZmpzEgM1UrXIpIy11kZvYszRcSA/rHLJF0a2bGxPxsFqzbQUODk6AbLkV6rdbGYG7v4GfSy5111BC++8QS/rKglIumFbTdQER6pBYLjLvP7cog0nOcP3kYj767np//fQWnjR9Ev4yUsCOJSAg6OgYj0qKEBOMn5x3Jruo6bntxZdhxRCQkKjASE4cP6cvVx45kVtF6Fq2vCDuOiIRABUZi5tunHUZeVirfe3oZ9Q268FCkt+nIVWQAuPu5MUkkPUZmahLfP2c8Nz2yiIfnr+PKY0eGHUlEulBrZzC3A3cQmRZmH3Bv8KgClsU+mvQEZx81hOPHDOD2F1dSvrs67Dgi0oVaLDDuPje4kmyyu1/s7s8Gj68Ax3VdROnOzIwfnXsE++sa+O/nVoQdR0S6UDRjMAPNbPSBN2Y2ChgYu0jS04wemMnXThzN04s3MW/19rDjiEgXiabAfAd4zcxeM7PXgFeBb8cylPQ8N558KAW5ffj+M8uoqWsIO46IdIFoput/ARhDZNnkbwFj3f3FWAeTniUtOZEffvEISsqr+OObWu1BpDdos8CYWTrwb8BN7r4EGG5m58Q8mfQ4px4+iNPHD+LXc1axsXJf2HFEJMai6SK7D6ghstAYQCnw05glkh7tB18cD8CPn30/5CQiEmvRFJhD3P02oBbA3feh9WCkg/L7pfONUw/lxffLeOWDsrDjiEgMRVNgasysD8FNl2Z2CLA/pqmkR7vuuNEcMjCDrz+8kJseWcicFWXU1mvgX6SniWbJ5B8CLwAFZvZn4HPAVTHMJD1cSlIC9189nXvfWMOzSzbxt6Wb6ZeezNkThnD+5GFMGd4PM50ki3R35t72HFFm1h+YQaRrbL67b4t1sK5QWFjoxcXFYcfo1WrrG3hj1VaeWrSJl5Zvobq2gYLcPsycOIzzJg/l0LyssCOKSBNmtsDdC9vcr60CY2ZzgDvc/flG2+5x9+sPPma4VGDiS9X+Ol5ctoWnF2/krZJtNDiMG5zF5w8fxCmH5zEpP0crZIrEgc4sMGuADcAr7v6jYNtCd5/SKUlDpAITv8p3V/Psks28+P4WFqyroL7BGZCZwklj8zh1XB7HjRlAVlpy2DFFeqXOLDALgenAr4EC4DLgVRUY6SqVe2uY++FW5qwo57WV5eyqriM50Zgxuj+njMvj84cPoiA3PeyYIr1GZxaYRe4+OXh9FfBdoJ+753dG0DCpwHQ/dfUNLFhXwZwPypmzoozVW/cAMHZQFqeNH8Rp4wdx1LBsdaWJxFBnFpivufvdjd5PBW5092sOPma4VGC6v7Xb9vDyijJeWl5G0dodNDgM6pvK5w+PFJtjDulPalJi2DFFepSDLjBm1tfdd5lZbnOfu/uOg8wYOhWYnqViTw2vfFDOS8vLeH3VVvbW1JOZmsSJhw3kxMMGMmVEPw4ZmKFLoEUOUmcUmL+5+zlm9hGRmywb/1/p7j662Yaf/o4zgDuBROAP7n5rk88t+PwsYC9wlbsvbK1tUPAeA0YCa4GL3L0i+GwCcDfQF2gAprl7i6tcqcD0XNW19by9ehsvLS/j5RXlbN0duTc4Jz2ZKcP7MWV4DlNG9GNSQQ7pKdHcDiYiB3RaF9lBBEgEPgROIzJ/WRFwqbsvb7TPWcA3iBSYo4E73f3o1tqa2W3ADne/1cxuITIedLOZJQELgcvdfUlw706lu9e3lFEFpndoaHDWbNvDwnUVLFhXwYL1FZSUVwGQmGAcPiSLqcP7cdK4PE4YM5BEjd+ItCraAtPiP93MrNWrxA6cabRiOlDi7muC75sFzASWN9pnJvCgR6rcfDPLMbMhRM5OWmo7EzgpaP8A8BpwM3A6sDSY8Rl318pWAkBCgnFoXiaH5mVy0bQCIHJl2qL1lSxcHyk6Tywo5YF56xiSncaFhQVcODVfV6aJHKTW+gbuaOUzB05p47uHEbl/5oBSImcpbe0zrI22g9x9M4C7bzazvGD7YYCb2YtEVtycFUzS+Slmdj1wPcDw4cPb+AnSU+Wkp3DyuDxOHhf5z6emroE5K8qYVbSB37yyit+8sorjDh3AJdOGc9r4QaQkRTNtn4g01mKBcfeTD/K7m+tnaNof19I+0bRtKgk4DphGZDxnTnAaN+dTX+J+D3APRLrI2vhO6SVSkhI486ghnHnUEDZW7uPxog08UbyBGx9ZSG5GChdMGcbF0wo0dY1IO0Q1umlmRwLjgbQD29z9wTaalRK5MfOAfGBTlPuktNK2zMyGBGcvQ4DyRt8198A8aWb2PDAF+FSBEWnLsJw+fOe0w/jmqWN4Y9VWHivawH1vreXeNz7i0LxMCkf0Y+qIfhSOzGVk/3RdlSbSgjYLjJn9F5Exj/HA88CZwJtAWwWmCBhjZqOAjcAlwFea7DMbuCkYYzka2BkUjq2ttJ0NXAncGjw/E2x/Efj3YAXOGuBE4Jdt/T6RliQmGCeNzeOksXlsq9rP04sic6Q9/95mZhVFenAHZKYwZXg/Ckf2Y+qIXI4c1lf33YgEojmD+TIwEVjk7leb2SDgD201cvc6M7uJyF/8icCf3P19M7sh+PwuIgXrLKCESLfW1a21Db76VuBxM7sWWA9cGLSpMLNfEClsDjzv7s9FcxBE2jIgM5Xrjh/NdcePpqHBKdlaRfHaCorX7WDBugr+sTyyeFpqUgInj83j8mNGcOwh/XV2I71aNHfyv+vu081sAXAysBtY5u5HdEXAWNJlytJZyndXs3BdBfPX7GD2kk3s2FPD6IEZXHb0CC6Ymk92H03MKT1HZ04V8zvg/xHppvouUAUsdverOyNomFRgJBaqa+v5+7LNPDRvHQvXV5KWnMB5k4Zx2YwRHDksO+x4IgctJjdamtlIoK+7Lz2IbHFDBUZibdnGnfz5nXU8vWgT+2rrmVSQw+UzRnD2hCGkJWusRrqnTi0wwRQsI2k0ZuPuTx5MwHigAiNdZee+Wp5cWMpD89exZuse+qYlcd7kYVxUWKCzGul2OrOL7E/ABOB9IvN7QWQuMs2mLNJO7s681duZVbSBF97fQk1dA0cO68vFhQWcO2mYxmqkW+jMArPc3cd3WrI4ogIjYarcW8MzizfxWNEGlm/eRWpSAmcdNYSLCguYMTpXV6BJ3DroucgamWdm4xtPUikiBy8nPYUrjx3JlceOZNnGnTxWtIGnF2/kqUUbGdE/nR+cM55TDx8UdkyRDovmDOYE4FlgC7CfyDQu7u4TYh8vtnQGI/HmwBVod89dw4dlu/nRuUdw+TEjw44l8imdeQbzJ+By4D0+GYMRkRhIS07k/Mn5fOGIwXzz0UV8/5n3Ka3cx81fGKdloKXbiabArHf32TFPIiIfS09J4u7LC/nh7Pe5e+4aNlbs4/YLJ+rSZulWoikwH5jZI0S6yfYf2NgTLlMWiWeJCcaPZx5Bfr8+/PzvH1C+ez/3XD6VnPSUsKOJRCWaRS76ECkspwNfDB7nxDKUiESYGV878RB+felkFq+v5ILfv82GHXvDjiUSlVbPYIKli7e5+791UR4Raca5E4cyKCuVrz5YzPm/e5v7rprGUfm6QVPiW6tnMMF69q0unSwiXePo0f158p+PJTUpgYvunscrH5SFHUmkVdGMwSw2s9nAE8CeAxs1BiPS9Q7Ny+KpG4/lmvuLuO6BYsYO7ktuRjL90lPIzUj55Dkjhdz0FPplJJOXlUb/jBRdhSZdLpoCkwtsB05ptM0BFRiREORlpfHY9cdw55xVrNm6h4q9NSzftIsde2uo3FvbbJvkRCMvK43B2WkM7pvGoL5pDMlOY1B25Hn0gAz6Z6Z28S+Rnq5dsyn3NLrRUnqauvoGdu6rpWJvDRV7a9letZ+yXfvZsquasp3VbNlVzZbgeW9N/afa9s9IYcygTMbkZXHYoEzGDMpiTF6mCo98RqfdaGlm+cBvgM8ROXN5E/iWu5cedEoR6VRJiQn0z0xtsyi4O7v311G2s5qNlfsoKa+ipLyKD8t28/SijezeX/fxvv0zUhg7OIuZk4Zy7sRh9EnRvTgSnWiminkJeAR4KNh0GfBP7n5ajLPFnM5gRD7L3dmyq5oPy6pYVbabD8t2s3B9JSXlVfRNS+LCwgIumzGCUQMywo4qIenM2ZQXu/uktrZ1RyowItFxd4rWVvDgvLW8sGwLdQ3O8WMGcPmMEZx6+CASdQFBr9KZc5FtM7PLgEeD95cSGfQXkV7CzJg+Kpfpo3Ip31XNrKINPPLOeq5/aAHDcvrwlaOHc1FhAQOzNF4jn4jmDGY48H/AMUTGYN4mMgazLvbxYktnMCIdV1ffwMsrynho/jreKtlOSmICV31uJDeefKgWTuvhOnXJ5J5KBUakc5SUV3HX3NX8dWEpOX2S+c5ph3Hp9OEkJ0YzG5V0NwddYMzsB620c3f/SUfDxQsVGJHOtWzjTn723ArmrdnOIQMz+M+zD+fksXlanbOHibbAtPbPiz3NPACuBW4+6IQi0uMcOSybR756NPdeUYg7XHN/MZf98R2Wb9oVdjQJQVRdZGaWBXyLSHF5HLjD3ctjnC3mdAYjEju19Q38ef46fjVnFTv31XLR1AK+e/ph5PVNCzuaHKTOOIPBzHLN7KfAUiJXnE1x95t7QnERkdhKTkzgqs+NYu6/nsy1nxvFk4tKOfWOuVpuoBdpscCY2f8CRcBu4Ch3/6G7V3RZMhHpEbLTk/neOeN57pvHs6emjllF68OOJF2ktTOY7wJDge8Bm8xsV/DYbWbqUBWRdjlsUBanjMvjieJS6uobwo4jXaDFAuPuCe7ex92z3L1vo0eWu/ftypAi0jNcPG045bv38+rKrWFHkS6gi9RFpMucPHYgeVmpPKZusl5BBUZEukxSYgJfnprPKx+Us2VnddhxJMZUYESkS108rYAGh78s2BB2FIkxFRgR6VIj+mdw7CH9eax4Aw0NvXeqqt5ABUZEutzF0wrYsGMf89ZoYvaeTAVGRLrcF44YTE56MrOK1E3Wk6nAiEiXS0tO5PzJw3hx2RZ27KkJO47EiAqMiITi4mkF1NQ38NSijWFHkRhRgRGRUIwb3JdJBTk8VrSe3rwuVU8W0wJjZmeY2UozKzGzW5r53Mzs18HnS81sSlttgwk4XzKzVcFzv2D7SDPbZ2aLg8ddsfxtInLwLp1ewIdlVSzaUBl2FImBmBUYM0sEfgucCYwHLjWz8U12OxMYEzyuB34fRdtbgDnuPgaYE7w/YLW7TwoeN8Tml4lIZzlnwlAyUhKZ9a7u7O+JYnkGMx0ocfc17l4DzAJmNtlnJvCgR8wHcsxsSBttZwIPBK8fAM6L4W8QkRjKSE3iixOH8uySzeyurg07jnSyWBaYYUDjaxBLg23R7NNa20HuvhkgeM5rtN8oM1tkZnPN7PiD/wkiEmsXTytgX209f1u6Oewo0sliWWCaW4S76UheS/tE07apzcBwd58M/AvwiJl9ZtZnM7vezIrNrHjrVs3oKhK2SQU5jBucpXtieqBYFphSoKDR+3xgU5T7tNa2LOhGI3guB3D3/e6+PXi9AFgNHNY0lLvf4+6F7l44cODADv40EeksZsbF0wpYsqGSFZu11FRPEssCUwSMMbNRZpYCXALMbrLPbOCK4GqyGcDOoNurtbazgSuD11cCzwCY2cDg4gDMbDSRCwfWxO7niUhnOX/yMFKSEnhMZzE9SswKjLvXATcBLwIrgMfd/X0zu8HMDlzh9TyRIlAC3Av8c2ttgza3AqeZ2SrgtOA9wAnAUjNbAvwFuMHdd8Tq94lI58lJT+GMIwbz5MJSqmvrw44jncR68w1OhYWFXlxcHHYMEQHeXr2Nr9z7DndeMomZk5peDyTxxMwWuHthW/vpTn4RiQszRvVnRP90HtU9MT2GCoyIxIWEhMhg//w1O7jugWLeXr1NU8h0c0lhBxAROeDa40ZRXVPPw++s5+V7yzh8SF+u+dxIzp00lNSkxLDjSTtpDEZjMCJxp7q2nmcWb+RPb65lZdluBmSmcNmMEVw2YwQDMlPDjtfrRTsGowKjAiMSt9ydt0q288c31/Dqyq2kJCVw3qShXHPcKMYN/sx91NJFoi0w6iITkbhlZhw3ZgDHjRnA6q1V3PfWR/x1wUYeLy7lixOH8m+nj2V4//SwY0oLdAajMxiRbqVybw1/eOMj/vDmGuobnH86egTfOOVQ+qvrrMuoiywKKjAi3VfZrmp+9fIqHi/eQJ/kRL52wmiuPX4U6SnqmIk1FZgoqMCIdH8l5VXc9sIH/GN5GQOzUvn258dwcWEBSYm6CyNWdKOliPQKh+Zlcs8Vhfz168cwIjed/3xqGaf/6nVeWLZF99GETAVGRHqEqSNyeeKGY7jn8qkYcMPDCzj/d28zb/X2sKP1WiowItJjmBmnHzGYF799ArddMIGyXdVceu98rvjTuyzbuDPseL2OxmA0BiPSY1XX1vPQvHX89rUSKvfWcs6EIfzr6WMZOSAj7Gjdmgb5o6ACI9I77Kqu5Z65a/jjmx9RW9/AxdMK+OapYxjUNy3saN2SCkwUVGBEepfy3dX83yslPPLOepISjetPOIRvnTqGxITmVmmXlugqMhGRJvKy0vjxzCOZ890TOW38YH49ZxX/+sQS6uobwo7WI+mOJBHpdUb0z+A3l05m7KBMbv/Hh9TUNfCrSyaRrHtnOpUKjIj0WjedMoa05ER++twK9tc18Nt/mqxlATqRyrWI9GrXHT+an8w8gpdXlHHdA8Xsq6kPO1KPoQIjIr3e5ceM5LYLJvBmyTauvv9d9uyvCztSj6ACIyICXDStgF9dPImitRVc/sd32FVdG3akbk8FRkQkMHPSMP7v0sm8t3Enl/3hHSr31oQdqVtTgRERaeTMo4Zw9+VT+WDLbi65Zz7bqvaHHanb0o2WutFSRJrx5qptXPdgEckJCQzN6UNuRgq5mSn0z0ihX3oK/TNTItvSUxjWrw/Dc9Mx6x03bGrJZBGRg3DcmAE8+tUZPF68ge1VNezYU8OKTbvYvqeGnfs+Oz7TPyOFKSP6MTV4HDUsm7Tk3n3JswqMiEgLJg/vx+Th/T6zva6+gYq9tezYU8P2Pfv5aNseFq6rZOH6Cl5aXgZAcqIxfmg2U4dHCs6YQZn0SU4kLTmRPimJpCUl9PhF0dRFpi4yEelE26v2s3B9pNgsWFfBkg2V7K9rfiqa5EQj7UDRSU5kYkEO//XF8QzITO3i1O2jLjIRkRD0z0zltPGDOG38IABq6hpYsXkX63fspbq2nuraevbV1lNd28C+2nr21dSzv66e3dV1vPj+Ft4u2catF0z4uH13pgIjIhJDKUkJTCzIYWJBTpv7rtyym28/tpivPljMJdMK+P4548lI7b5/TffsDkARkW5k7OAsnr7xWG448RAeK97AmXe+wYJ1FWHH6jAVGBGROJKalMgtZ47jseuPocGdC+96mzv+sZLabrikgAqMiEgcmj4ql79/63i+NCWf37xSwpd+9zYl5bvDjtUuKjAiInEqKy2Z2y+cyF2XTaG0Yi9n//pNbn9xJe+V7qShIf6vANZlyrpMWUS6gfLd1fznU8t4eUUZ7jAwK5WTxw7klHGDOG7MADK78GKAaC9TVoFRgRGRbmRb1X7mrtzKKx+U8/qHW9m9v47kROPoUf05eVwep47LY+SAjJhmUIGJggqMiHRntfUNFK+t4NWV5bzyQTkl5VUAFOT2YdrIXKaPzGX6qFxGDcjo1HnSVGCioAIjIj3J+u17eXVlOfNWb6do7Q6274ksNzAgM4VpI3MjRWdULocP6UtiQscLjgpMFFRgRKSncndWb91D0dodFH20g3fX7qC0Yh8AWalJXDytgO+dM75D362pYkREejEz49C8TA7Ny+TS6cMB2FS5j6K1O3j3ox0MzekT8wwxvUzZzM4ws5VmVmJmtzTzuZnZr4PPl5rZlLbamlmumb1kZquC535NvnO4mVWZ2b/G8reJiHQ3Q3P6MHPSMH52/lFcc9yomP95MSswZpYI/BY4ExgPXGpmTc/HzgTGBI/rgd9H0fYWYI67jwHmBO8b+yXw907/QSIi0i6xPIOZDpS4+xp3rwFmATOb7DMTeNAj5gM5ZjakjbYzgQeC1w8A5x34MjM7D1gDvB+bnyQiItGKZYEZBmxo9L402BbNPq21HeTumwGC5zwAM8sAbgZ+1FooM7vezIrNrHjr1q3t+kEiIhK9WBaY5q6Ba3rJWkv7RNO2qR8Bv3T3qtZ2cvd73L3Q3QsHDhzYxleKiEhHxfIqslKgoNH7fGBTlPuktNK2zMyGuPvmoDutPNh+NPBlM7sNyAEazKza3f+vM36MiIi0TyzPYIqAMWY2ysxSgEuA2U32mQ1cEVxNNgPYGXR7tdZ2NnBl8PpK4BkAdz/e3Ue6+0jgV8B/q7iIiIQnZmcw7l5nZjcBLwKJwJ/c/X0zuyH4/C7geeAsoATYC1zdWtvgq28FHjeza4H1wIWx+g0iItJxupNfd/KLiLSLpoqJgpltBdYdxFcMALZ1Upyu0N3ygjJ3le6WubvlhZ6VeYS7t3mVVK8uMAfLzIqjqeLxorvlBWXuKt0tc3fLC70zs1a0FBGRmFCBERGRmFCBOTj3hB2gnbpbXlDmrtLdMne3vNALM2sMRkREYkJnMCIiEhMqMCIiEhMqMC0wszQze9fMlpjZ+2b2o2D7Y2a2OHisNbPFwfaRZrav0Wd3xUneSWY2P8hUbGbTG7X5j2BBt5Vm9oWuzNuRzGEf4zYyTzSzeWb2npk9a2Z9G7WJ1+PcbOZ4OM5BjkQzW2Rmfwvet7jYYNjHuL2Z4+UYt5D5wuC/kwYzK2yyb/uOs7vr0cyDyIzOmcHrZOAdYEaTfe4AfhC8Hgksi7e8wD+AM4PtZwGvBa/HA0uAVGAUsBpIjPPMoR7jNjIXAScG268BftINjnNLmUM/zkGOfwEeAf4WvL8NuCV4fQvwP/FyjDuQOS6OcQuZDwfGAq8BhY32a/dx1hlMCzziwNT/ycHj4ysizMyAi4BHQ4j3Ga3kdeDAv6az+WRW6pnALHff7+4fEZkPbjpdqAOZQ9dK5rHA68H2l4ALgtfxfJxbyhw6M8sHzgb+0GhzS4sNhn6Mod2Z40Jzmd19hbuvbGb3dh9nFZhWBKeOi4ksCfCSu7/T6OPjgTJ3X9Vo26jgVHOumR3flVmhxbzfBv7XzDYAtwP/EewezYJwMdfOzBDyMYYWMy8Dzg12uZBPlpuI5+PcUmYI/zj/Cvh3oKHRtmYXGyROjjHtywzhH2NoPnNL2n2cVWBa4e717j6JyHo0083syEYfX8qnz142A8PdfTLBKWfjfviu0ELerwPfcfcC4DvAH4PdO7KoW6drZ+bQj3Erma8BbjSzBUAWUBPsHs/HuaXMoR5nMzsHKHf3BdE2aWZblx7jDmQO/b/lrjjOKjBRcPdKIv2RZwCYWRLwJeCxRvvsd/ftwesFRPonD+vqrMGfX8knea8Engw+eoJPTmmjWRCuy0STOZ6OcZChkiCzu3/g7qe7+1Qi//BYHewWt8e5pcxxcJw/B5xrZmuBWcApZvYwwWKDAPbpxQbj4Ri3K3McHOPWMrek/cc5zMGleH4AA4Gc4HUf4A3gnOD9GcDcZvZPDF6PBjYCuWHnBVYAJwXbTwUWBK+P4NMDdmvo+sHn9mYO9Ri3kTkv2JYAPAhc0w2Oc0uZQz/OjbKfxCeDz//LpwfMb4uXY9yBzHFzjJtmbrTtNT49yN/u4xzLJZO7uyHAA2aWSOR/wMfd/W/BZ5fw2cH9E4Afm1kdUA/c4O47uixtC3nNrBK4MzjrqgauB/DI4m+PA8uBOuBGd6/vwrztzkz4x7i1zN8ysxuDfZ4E7oO4P87NZiY+jnNzml1sME6OcUtaWiAxXo8xZnY+8BsiRfA5M1vs7l/oyHHWVDEiIhITGoMREZGYUIEREZGYUIEREZGYUIEREZGYUIEREZGYUIEROQhmVtX2XuHrLjmlZ1GBEelGgnuDRLoFFRiRTmafrGez1MyearQGyDfNbHmwfVaw7cRGa4IsMrOsZr7vfjP7hZm9CvyPmU03s7eD/d82s7HBfleZ2ZNm9oJF1h+5rZnvGmCRNWDOjvFhENGNliIHw8yq3D2zybalwDfcfa6Z/Rjo6+7fNrNNwCh3329mOe5eaWbPAre6+1tmlglUu3tdk++7HxgAzHT3+mBSxL3uXmdmnwe+7u4XmNlVwA+AycB+YCVwnLtvCLrIDgFmA99z95dieFhEAJ3BiHQqM8smMu/X3GDTA0SmBQFYCvzZzC4jMtUGwFvAL8zsm0G7TxWXRp5oNC1HNvCEmS0DfklkjqgD5rj7TnevJjKlx4hgezIwB/h3FRfpKiowIl3nbOC3wFRggZklufutwHVEJqGcb2bjzOxnB7rNGrXd0+j1T4BX3f1I4ItAWqPP9jd6XQ8fzzdYBywAQltOWHofFRiRTuTuO4GKRgtIXQ7MNbMEoMDdXyWywFMOkGlmh7j7e+7+P0AxMM7d/9PdJ3lk/ZbmZBOZfRfgqmijEVn/ZZyZ3dLe3yXSEboiReTgpJtZaaP3vyCyns1dZpZOZErzq4FE4OGgC82AXwZjMD8xs5OJnG0sB/4exZ95G5HZkf8FeCXaoMH4zSXAs2a2y91/F21bkY7QIL+IiMSEushERCQmVGBERCQmVGBERCQmVGBERCQmVGBERCQmVGBERCQmVGBERCQm/j+KmU1dEBrOkwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[375:410]\n",
    "x_axis = np.arange(375,410)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "id": "6dc089a7",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[395, 0.9722222222222222, 0.8536585365853658, 0.9090909090909091, 0.9746543778801844, 0.9956845821890937, 0.9850572482049291, 0.947074078647919]\n",
      "false positives 11\n"
     ]
    }
   ],
   "source": [
    "j= 395\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f5f84615",
   "metadata": {},
   "source": [
    "GAT huge 15%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "e46113ab",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\huge-fifteen\\huge-0.1-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\huge-fifteen\\huge15besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "fb449b7c",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "b0da0467",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 132\n",
      "[363, 0.967032967032967, 0.7822222222222223, 0.8648648648648648, 0.9628224582701063, 0.9952941176470588, 0.978789047435403, 0.9218269561501339]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "5f1e05ae",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "bca9006d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[300:400]\n",
    "x_axis = np.arange(300,400)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "id": "4d377c5c",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[350:370]\n",
    "x_axis = np.arange(350,370)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "7a790c64",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[365, 0.9617486338797814, 0.7822222222222223, 0.8627450980392157, 0.9627942293090357, 0.9945098039215686, 0.9783950617283952, 0.9205700798838055]\n",
      "false positives 14\n"
     ]
    }
   ],
   "source": [
    "j= 365\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "7fd10280",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[357, 0.9692737430167597, 0.7711111111111111, 0.8589108910891089, 0.9610143830431491, 0.995686274509804, 0.9780431432973806, 0.9184770171932448]\n"
     ]
    }
   ],
   "source": [
    "j= 357\n",
    "search_more = True\n",
    "while search_more:    \n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if (false_positives/positives > 0.03):\n",
    "        search_more = False\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    best_score = f1_score\n",
    "    precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "    recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "    f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "    macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "    best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                f1_score_on_normal, macro_f1_score]\n",
    "    j += 1   \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0cfcd2a4",
   "metadata": {},
   "source": [
    "GAT small 15%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "a5f8b13b",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\small-fifteen\\small-0.1-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\small-fifteen\\small15besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "301d2282",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "79f639ad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 385\n",
      "[386, 0.9974160206718347, 0.8301075268817204, 0.9061032863849765, 0.969766551856104, 0.9996055226824457, 0.9844599844599843, 0.9452816354224804]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "b4a300fc",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "a5a29240",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[300:400]\n",
    "x_axis = np.arange(300,400)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "5238a37f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZMAAAEGCAYAAACgt3iRAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjQuMywgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/MnkTPAAAACXBIWXMAAAsTAAALEwEAmpwYAAAor0lEQVR4nO3deXidZZ3/8fc3e9IkTdOkW7okpS2lZSklDagssgoV6bghqIiAw1QFER0VRy9n1PmNiAsjI4soq4oIw1YcVhFaQWibllLaQtvQfU3SvUmzf39/nCcQ0iwnPTk5Sz6v6zrXOec5933O9/aRfvJs92PujoiISCRSYl2AiIgkPoWJiIhETGEiIiIRU5iIiEjEFCYiIhKxtFgXMBCKioq8tLQ01mWIiCSUJUuW1Lp7cThtB0WYlJaWUllZGesyREQSipltDLetdnOJiEjEFCYiIhIxhYmIiERMYSIiIhFTmIiISMQUJiIiEjGFiYiIRExh0oMFa2q47aWqWJchIhL3FCY9eKWqll8+t4Y9dU2xLkVEJK4pTHpw0YwxtLQ5T63YHutSRETimsKkB9NG5zNpRC5PvL4t1qWIiMQ1hUkPzIw5J4xh0YbdbN17KNbliIjELYVJL+bMKAHgyTe0dSIi0h2FSS/GD8/hxPEFPLFMYSIi0h2FSRjmnDCGt7bvZ83OA7EuRUQkLilMwvDR48eQmmI8sWxrrEsREYlLCpMwFOdl8qFJRTyxbBvuHutyRETijsIkTHNOGMOWPYdYumlvrEsREYk7UQ0TMzvfzFabWZWZ3dDF52ZmtwSfLzezmR0+u9vMqs1sRac+hWb2vJmtDZ6HRXMM7c6bPpLMtBTmaVeXiMhhohYmZpYK3ApcAEwDLjWzaZ2aXQBMDh5XA7d3+Oxe4PwuvvoG4AV3nwy8ELyPurysdM45ZiR/Wb6d5ta2gfhJEZGEEc0tkwqgyt3XuXsT8CAwp1ObOcD9HvIaUGBmowHcfQGwu4vvnQPcF7y+D/inaBTflYtmjGFXXROvVNUO1E+KiCSEaIZJCbC5w/stwbK+tulspLtvBwieR3TVyMyuNrNKM6usqanpU+Hd+fDRxeRnpTFP15yIiLxPNMPEuljW+VSocNocEXe/093L3b28uLi4P76SzLRUZh83mmdX7uBQU2u/fKeISDKIZphsAcZ1eD8W6PwnfThtOtvZvisseK6OsM4+uWjGGOqaWvnrWzsH8mdFROJaNMNkMTDZzMrMLAO4BJjXqc084AvBWV2nAPvad2H1YB5wefD6cuCJ/iy6NyeXDWdkfqamVxER6SBqYeLuLcA1wLPAW8BD7r7SzOaa2dyg2VPAOqAK+C3wlfb+ZvYn4FXgaDPbYmZXBR/dCJxrZmuBc4P3AyY1xfjY8WOYv6aavfW6aZaICIANhiu6y8vLvbKyst++b8XWfVz4Py/zk08cx6UV4/vte0VE4omZLXH38nDa6gr4IzB9TD4Ti4dori4RkYDC5AiEbppVwsL1u9m+TzfNEhFRmByhi2aMwV03zRIRAYXJESsrGsIJY4fqrC4RERQmEZkzo4SV2/ZTVa2bZonI4KYwicCFx48mxdD0KiIy6ClMIjAiP4sPHlXE47pplogMcgqTCF00YwybdtezbPPeWJciIhIzCpMInX/sKDLSUnQgXkQGNYVJhPKz0jl76gj+snw7LbpplogMUgqTfjBnxhhqDzbyj3d2xboUEZGYUJj0gw8fPYKMtBQWrOmfm3CJiCQahUk/yEpPZcbYAhZv6OouwyIiyU9h0k8qygpZsW0/BxtbYl2KiMiAU5j0k4qyQlrbnKUb98S6FBGRAacw6SczJwwjxdCuLhEZlBQm/SQ3M41jS4aycL3CREQGH4VJP6ooLWTZ5r00NLfGuhQRkQGlMOlHFWWFNLW0sXzLvliXIiIyoBQm/WhWaSGg4yYiMvgoTPrRsCEZTBmZq+MmIjLoKEz6WUVZIUs27NY8XSIyqChM+tms0kLqmlp5a7vuvigig4fCpJ9VlIWOmyxcr0kfRWTwUJj0s9FDsxlfmMMiHTcRkUFEYRIFFWWFLN6wW7fyFZFBQ2ESBRWlheypb6aq+mCsSxERGRAKkyh477iJdnWJyOCgMImCCcNzGJGXqeMmIjJoKEyiwMyYVVbIovU6biIig4PCJEpOLitkx/4Gtuw5FOtSRESiLqphYmbnm9lqM6sysxu6+NzM7Jbg8+VmNrO3vmY2w8xeM7NlZlZpZhXRHMOR0nETERlMohYmZpYK3ApcAEwDLjWzaZ2aXQBMDh5XA7eH0fcm4IfuPgP4QfA+7kwZkcfQ7HQWK0xEZBCI5pZJBVDl7uvcvQl4EJjTqc0c4H4PeQ0oMLPRvfR1ID94PRTYFsUxHLGUFGNW6TAWaQZhERkEohkmJcDmDu+3BMvCadNT368DPzOzzcDPge929eNmdnWwG6yypqbmSMcQkYqyQtbX1lF9oCEmvy8iMlCiGSbWxbLOpzZ116anvl8Grnf3ccD1wF1d/bi73+nu5e5eXlxcHGbJ/auibDgAi9fvicnvi4gMlGiGyRZgXIf3Yzl8l1R3bXrqeznwaPD6YUK7xOLS9DH5ZKenskiTPopIkotmmCwGJptZmZllAJcA8zq1mQd8ITir6xRgn7tv76XvNuCM4PVZwNoojiEi6akpnDRhmM7oEpGklxatL3b3FjO7BngWSAXudveVZjY3+PwO4ClgNlAF1ANX9NQ3+Op/Bn5lZmlAA6GzwOJWRVkhN/91Dfvqmxmakx7rckREoiJqYQLg7k8RCoyOy+7o8NqBr4bbN1j+MnBS/1YaPRVlhbhD5cbdnH3MyFiXIyISFboCPspmjCsgPdU0T5eIJLU+hYmZpZhZfu8tpV1WeionjC3QcRMRSWq9homZPWBm+WY2BFgFrDazb0W/tORRUVbIiq37qG9qiXUpIiJREc6WyTR33w/8E6FjGOOBy6JZVLKpKCukpc15fdPeWJciIhIV4YRJupmlEwqTJ9y9mcMvPpQenDRhGCmmSR9FJHmFEya/ATYAQ4AFZjYB2B/NopJNXlY608bka9JHEUlavYaJu9/i7iXuPjuYkHEjcOYA1JZUKkqHs3TTHppa2mJdiohIvwvnAPx1wQF4M7O7zGwpoSvPpQ8qyobR2NLGm1v3xroUEZF+F85uriuDA/DnAcWErlK/MapVJaFZpaGbZS3SpI8ikoTCCZP2GXxnA/e4+xt0Pauv9GB4biaTRuRq0kcRSUrhhMkSM3uOUJg8a2Z5gHb8H4GKskIqN+yhtU0nw4lIcgknTK4CbgBmuXs9kEEwIaP0TUVpIQcaW3hru06GE5Hk0utEj+7eZmZjgc+aGcB8d38y6pUloYqy0HGTxRt2c2zJ0BhXIyLSf8I5m+tG4DpCU6msAr5mZj+JdmHJaExBNmOHZWvSRxFJOuHs5poNnOvud7v73cD5wEejW1byOmXicF5cXc3f18bmvvQiItEQ7qzBBR1ea/9MBL71kaMpHT6EK+5ZzCNLtsS6HBGRfhFOmPwEeN3M7jWz+4AlwH9Ft6zkNTI/i4fmfoCTJxbyzYff4Nd/W0voHmEiIokrnOlU/gScAjwaPD4ArI9yXUktPyude75YwT/NGMPPn1vD9x5fQUurzrYWkcQV1m173X07MK/9vZktIjQVvRyhjLQUbv7MDEYXZHP7S+9Qvb+BWy49kZyMqN5JWUQkKo70tr26Ar4fmBnfOX8qP54znb+9Xc2lv11I7cHGWJclItJnRxom2snfjy77QCl3fP4kVu/Yzydv/wcbautiXZKISJ9Ydwd/zexJug4NA85y9yHRLKw/lZeXe2VlZazL6NXSTXv40n2hOu+6vJwTxw+LcUUiMpiZ2RJ3Lw+rbQ9hckZPHd19/hHUFhOJEiYA62vruPzuRVQfaODXl87knGkjY12SiAxS/RImySSRwgSg9mAjV927mDe37uNjJ4zh4yeWcOqkItJSj3SvpIhI3/UlTHTqUBwqys3kT1efwk3PrOax17fyxLJtFOVmclEQLMeW5BPMkyYiEhe0ZRLnGltaeWl1DY8t3crf3q6mqbWNSSNy+fiJJcyZMYaxw3JiXaKIJCnt5uokkcOko331zfzfm9t5/PWtLNoQmiyyoqyQT5xYwuzjR5OflR7jCkUkmfTXAfjuzuYCwN0vOrLyBl6yhElHm3fX88SyrTz6+lbW1dQxYXgOz11/OplpqbEuTUSSRF/CpKcjuj8HfkFo6pRDwG+Dx0FgRaRFSmTGFeZwzVmTeeEbZ3DrZ2eycVc9Dy3eHOuyRGSQ6vYAfPupv2b2Y3c/vcNHT5rZgqhXJmExM2YfN4ryCcO47aV3uHjWOG2diMiAC+dc02Izm9j+xszKgOJwvtzMzjez1WZWZWY3dPG5mdktwefLzWxmOH3N7Nrgs5VmdlM4tSQzM+O6cyazfV8DD1dqWnsRGXjhnBp8PfCSma0L3pcC/9JbJzNLBW4FzgW2AIvNbJ67r+rQ7AJgcvA4GbgdOLmnvmZ2JjAHON7dG81sRBhjSHqnTipi5vgCbnuxik+Xj9XWiYgMqHCmoH+G0D/21wWPo9392TC+uwKocvd17t4EPEgoBDqaA9zvIa8BBWY2upe+XwZudPfGoL7qMGpJembG18+ZwrZ9DfyvbrolIgMsnHvA5wDfAq5x9zeA8WZ2YRjfXQJ0PCK8JVgWTpue+k4BTjOzhWY238xmdVP31WZWaWaVNTWD4xa5p00u4sTxBdz24js0tej+KCIycMI5ZnIP0EToplgQ+of9P8Po19Ul2p1PNe6uTU9904BhhG7Y9S3gIevicnB3v9Pdy929vLg4rEM8Cc/MuO7syWzde0hbJyIyoMIJk6Pc/SagGcDdDxHe/Uy2AOM6vB8LbAuzTU99twCPBrvGFgFtQFEY9QwKZ0wpZsa4Am59sUpbJyIyYMIJkyYzyybYMjCzo4Bw7uC0GJhsZmVmlgFcQoe7NQbmAV8Izuo6BdgX3NWxp76PA2cFtUwBMoDaMOoZFNrP7Nq69xCPLtXWiYgMjHDC5D+AZ4BxZvZH4AXg2711cvcW4BrgWeAt4CF3X2lmc81sbtDsKWAdUEXogsiv9NQ36HM3MNHMVhA6MH+5D4Y5Yfrgw1OKOWHsUH79YhXNure8iAyAsObmMrPhhI5RGPCauyfUlkAyTqfSmxffruaKexfz008ex2dmjY91OSKSgPprOpX2L3sBONnd/8/d/+LutWZ2Z8RVSlR9+OhijtfWiYgMkHB2c5UB3zGzf++wLKykkthpP7Nr8+5DPLZ0a6zLEZEkF06Y7AXOBkaa2ZNmNjS6JUl/OWvqCI4r0daJiERfOGFi7t7i7l8BHgFeBjSFSQJo3zrZtLuex17X1omIRE84YXJH+wt3vxf4IvBclOqRfnb2MSM4tiSfW1+sokVbJyISJd2GiZnlBy8fNrPC9geh+5v864BUJxEzM7521mQ27qrn8WWdrxkVEekfPW2ZPBA8LwEqg+clHd5Lgjh32kimjc7nf/62VlsnIhIV3YaJu18YPJe5+8Tguf0xsbt+En/ar4rfuKueJ7R1IiJR0O39TDreqKor7r60/8uRaDlv2kiOGZ3Pr1+sYs6MMaSlhnO4TEQkPD3dHOsXPXzmBPNjSWJoP7Nr7h+W8M2H3+B7Hz2GEXlZsS5LRJJET/eAP3MgC5HoO2/aSK45cxK/WfAOf3urmuvOmczlHywlXVspIhKhcOfmOhaYBrz7p6y73x/FuvrVYJybqyfrag7ywydXMX9NDZNG5PLDi6bzoUmaxV9E3q+/5+b6d+B/gseZwE3ARRFVKDE1sTiXe6+YxW+/UE5TSxuf+91CvvyHJWzZUx/r0kQkQYWzf+NThKZT2eHuVwAnAJlRrUqizsw4d9pInrv+dL557hReXF3NOb+czy0vrKWhuTXW5YlIggknTA65exvQElzIWA3o1OAkkZWeyrVnT+aFb36Ys6eO5JfPr+Hcm+fz3Mod6DYxIhKucMKk0swKCN28agmwFFgUzaJk4JUUZHPr52bywJdOJjs9lat/v4S5f1hCY4u2UkSkd2EdgH+3sVkpkO/uy6NWURToAHzfNLe2cdfL67nx6be54NhR/PqzM0lNsViXJSIDrF8PwAdfeLyZXQTMBCaZ2SciKVDiW3pqCnPPOIrvf/QYnl6xg3979E3t8hKRHvV00SIAZnY3cDywEmif2MmBR6NYl8SBL502kb31zfz6xSoKhqTz3QuOiXVJIhKneg0T4BR3nxb1SiQuffO8Keypb+I389cxLCeDuWccFeuSRCQOhRMmr5rZNHdfFfVqJO6YGT+acyz7DjVz49NvU5CdziUV42NdlojEmXDC5D5CgbIDaAQMcHc/PqqVSdxITTF+efEMDjS08G+PvcnQ7HQuOG50rMsSkTgSTpjcDVwGvMl7x0xkkMlIS+H2z8/ksrsWcd2Dy8jLSufUyZqCRURCwjmba5O7z3P39e6+sf0R9cok7uRkpHH35bOYWDyEq39fyeub9sS6JBGJE+GEydtm9oCZXWpmn2h/RL0yiUtDc9K5/8oKinIzueLexazZeSDWJYlIHAgnTLIJHSs5D/hY8LgwmkVJfBuRn8UfrjqZjNQULrtrIZt3a4JIkcGuxzAxs1Sg1t2v6PS4coDqkzg1fngO919VwaGmVi67ayG1BxtjXZKIxFCPYeLurYSuehc5zNRR+dxzRQXb9zXwjYfeoK1NV8mLDFbh7OZaZmbzzOwyHTORzk6aMIzvXziNBWtquOcfG2JdjojESDinBhcCu3j/Pd81nYq86/Mnj2f+6hp++vTbnDKxkOljhsa6JBEZYH2aNThRadbg6Ntd18T5/72A/Ox0nrzmVLIzUmNdkohEqL9v2zvWzB4zs2oz22lmj5jZ2DALOd/MVptZlZnd0MXnZma3BJ8vN7OZfej7r2bmZqYr5+JA4ZAMfnHxCVRVH+Q//08z74gMNuEcM7kHmAeMAUqAJ4NlPQrOBLsVuACYBlxqZp0njLwAmBw8rgZuD6evmY0DzgU2hVG/DJDTJhdz9ekT+ePCTTy3ckesyxGRARROmBS7+z3u3hI87gWKw+hXAVS5+zp3bwIeBOZ0ajMHuN9DXgMKzGx0GH1vBr5N6NiNxJF/Pe9oji3J5zuPLGfn/oZYlyMiAyScMKk1s8+bWWrw+DyhA/K9KQE2d3i/JVgWTptu+wY36drq7m/09ONmdrWZVZpZZU1NTRjlSn/ISEvhV5ecSENzG994aJlOFxYZJMIJkyuBi4EdwHbgU8Gy3nR1n9fO/7J016bL5WaWA3wP+EFvP+7ud7p7ubuXFxeHsyEl/eWo4lz+/WPTeKVqF7/9+7pYlyMiA6DXU4PdfRNw0RF89xZgXIf3Y4FtYbbJ6Gb5UUAZ8IaZtS9famYV7q6d9HHkM7PG8dLqGn7+3Go+eFQRx43V6cIiyazbMDGznv76d3f/cS/fvRiYbGZlwFbgEuCzndrMA64xsweBk4F97r7dzGq66uvuK4ERHWrcAJS7e20vtcgAMzNu/ORxnP/fe/nag6/zl2tPZUhmOJc1iUgi6mk3V10XD4CrgO/09sXu3gJcAzwLvAU85O4rzWyumc0Nmj0FrAOqgN8CX+mpb9+GJrFWkJPBzZ+ZwYZddfzoSZ0uLJLMwrpo0czygOsIBclDwC/cvTrKtfUbXbQYWzc98za3vfQOt31uJrN1h0aRhNFvFy2aWaGZ/SewnNAusZnu/p1EChKJvevPncIJY4dywyPL2bb3UKzLEZEo6DZMzOxnhI57HACOc/f/cHfdWk/6LD01dLpwS5sz9w9L2KpAEUk6PW2ZfJPQVe/fB7aZ2f7gccDM9g9MeZIsSouGcPNnZvBO9UHOv3kBjyzZwmCYF05ksOg2TNw9xd2z3T3P3fM7PPLcPX8gi5Tk8JHpo3j6utOZOjqPbz78Bl/+w1J21zXFuiwR6QfhXLQo0m/GD8/hwas/wA0XTOWFt3dy3s0LePFtHYITSXQKExlwqSnG3DOO4omvnkpRbgZX3LuYf3vsTeoaW2JdmogcIYWJxMy0Mfk8cc2H+JfTJ/KnRZuYfcvfWbJR53iIJCKFicRUZloq3519DA/+8ym0tjmfvuMf/OzZt2lqaYt1aSLSBwoTiQsnTxzO09edxqdOGsutL77Dx297hR37NIW9SKJQmEjcyMtK56ZPncCdl51EVfVBfvbs6liXJCJhUphI3Dlv+ig+d/IEHl+2lU276mNdjoiEQWEicelfzphIaopx+/yqWJciImFQmEhcGpmfxWfKx/G/S7Zo+hWRBKAwkbg198NHAXDHS+/EuBIR6Y3CROJWSUE2nzppLH9evFlndonEOYWJxLWvfHgSre78ZoG2TkTimcJE4tq4whw+fmIJDyzcRPUBbZ2IxCuFicS9r545iebWNn739/WxLkVEuqEwkbhXVjSEi04Yw+9f3ciug42xLkdEuqAwkYRwzVmTaGhp5a6XtXUiEo8UJpIQJo3IY/Zxo7n/1Y3srdcNtUTijcJEEsa1Z03iYGMLd7+yIdaliEgnChNJGFNH5fOR6SO555X17G9ojnU5ItKBwkQSyrVnTeZAQwv3aetEJK4oTCShHFsylLOnjuCuV9ZzULf5FYkbChNJONeePZm99c38/tWNsS5FRAIKE0k4M8YVcPqUYn7393XUN2nrRCQeKEwkIX3trEnsqmvigYWbYl2KiKAwkQRVXlrIByYO5zcL1tHQ3BrrckQGPYWJJKyvnT2ZmgON3P/qBlpa22JdjsiglhbrAkSO1CkTC6koLeS/nnqbG59+m1H5WZQMy6akIDt4znnvfUE22RmpsS5ZJGlFNUzM7HzgV0Aq8Dt3v7HT5xZ8PhuoB77o7kt76mtmPwM+BjQB7wBXuPveaI5D4pOZcecXTuKZFTvYuvcQW/ccYsveQyzesIcnl2+ntc3f174oN4PyCYWccXQxp08ppqQgO0aViyQfc/feWx3JF5ulAmuAc4EtwGLgUndf1aHNbOBaQmFyMvArdz+5p75mdh7wN3dvMbOfArj7d3qqpby83CsrK/t9jBK/Wlrb2Hmgka17DrF1bz1b9xxifW09/3inlu3BXRsnj8jl9CnFnDGlmIqyQrLSteUi0pGZLXH38nDaRnPLpAKocvd1QVEPAnOAVR3azAHu91CivWZmBWY2Gijtrq+7P9eh/2vAp6I4BklQaakp7+7egsJ3l7s7a6sPsmBNDfPX1PD7Vzdy18vryUpP4ZSJwzkjCJeyoiGENpxFJBzRDJMSYHOH91sIbX301qYkzL4AVwJ/7urHzexq4GqA8ePH96VuSWJmxpSReUwZmceXTptIfVMLC9ftZn4QLj98MvS3TuGQDEblZzFqaBYj87MYPTSLUflZjAyeRw3NIj8rTYEjEohmmHT1X1nnfWrdtem1r5l9D2gB/tjVj7v7ncCdENrN1VuxMjjlZKRx5tQRnDl1BACbdtUzf001q7bvZ8e+Brbva2DZ5r3srjt82vvs9FRGF2Rx6qQiLjh2NBVlhaSmKFxkcIpmmGwBxnV4PxbYFmabjJ76mtnlwIXA2R6tgz4yKI0fnsNlHyg9bHlDcyvV+xvZsb+BHfsb2Lkv9LxxVx0PVW7m/lc3UpSbwXnTRzH72NGcMrGQtFSdeS+DRzTDZDEw2czKgK3AJcBnO7WZB1wTHBM5Gdjn7tvNrKa7vsFZXt8BznD3+ijWL/KurPRUxg/PYfzwnMM+q29q4aXVNTz15nYef30rDyzcxLCcdD4yfRQXHDeaDx41nHQFiyS5qIVJcLbVNcCzhE7vvdvdV5rZ3ODzO4CnCJ3JVUXo1OAreuobfPWvgUzg+WB/9WvuPjda4xDpTU5GGrOPG83s40ZzqKmV+WtqeHrFdv6yfDsPLt7M0Ox0zp02kguPH81pk4u1K0ySUtRODY4nOjVYYqGhuZWX19by1IrtPL9qJwcaWhg9NItPl4/j4vKxjB12+FaOSDzpy6nBChORAdDY0soLb1Xz4OLN/H1tDQCnTiriklnjOXfaSDLStBtM4o/CpBOFicSTLXvqebhyCw9XbmbbvgYKh2TwyZklfGbWOCaNyIt1eSLvUph0ojCReNTa5vx9bQ1/XryZ51ftpKXNKZ8wjE+Xj2XqqHyG52ZQlJupK/MlZhQmnShMJN7VHGjksde38ODizayrqXvfZ7mZaQzPzWD4kFC4DM/NpCgImoKcdPKz0snLSiMvK53crDTystLIzUgjRQf6JUIKk04UJpIo3J23th9g+75D1B5spPZgE7sONlF7sJFddY3UHmhiV10ju+uaaOvhP10zyM1Iezdc8rPSGTssm9KiIZQFj9KiIeRnpQ/c4CThxMvcXCLSR2bGtDH5TBuT32O71jZnT30Te+ub2N/QwsGGFg40tHCgofm958b3lu2tb2bR+t08vuz91w0PH5JBadEQSocPoawoh9KiIcwYV6AzzaTPFCYiCSg1xSjKzaQoN7NP/RqaW9m4q571tXVs2FXHhto61tfW8XJVDY8sbQRCWzXnHjOSq04to6KsUPOPSVgUJiKDSFZ6KkePyuPoUYefNVbf1ML62jqefnMHf1y4kedW7WT6mHyuOrWMC48fo9OXpUc6ZiIih2lobuWx17dy98vrWVt9kOK8TL5wygQ+d8oECodkxLo8GSA6AN+JwkTkyLg7C9bWcvfL65m/pobMtBQ+MbOEKz9UxuSRuiYm2ekAvIj0CzN794Zha3ce4O5XNvDo0i38adFmPjRpOBOLcslKTyE7PZWsjNTQc3qH54xUstJSyM9Opzgvk2E5GZqbLElpy0RE+mR3XRN/WrSJR5ZuYW99M4eaWmloaSWcf0pSU4zhQzIozssMPXJDzyPyMinOy2J4bgYZaSmkmJFqRkpKqE+KhR6pKaHlZpCTkcqwnAxdTxNF2s3VicJEJLrcncaWNhqaWznU3EpDcxuHmtpft7LvUDO1BxupOdBI9f5GaoLXNQcaqT3YSEtPF830IMWgcMh7F3GGLu7MpCgvg6IhwfvcTEbmh858060A+ka7uURkQJkZWcGurYI+9m0LrpmpOdjIroNNNLW24e60toWup2nz0KP9dWsbtLlT19jC7rqmDhd3NrJpUz27DjZS19TaRY0wfEgoWEbmZzEyP5MReVnvvg49Z1GUm6HToY+AwkREYiolxRgeTBPTXw41tYZmDDjYRO2BRqoPNLJzfwPVBxrYuT/0evmWfeyqazxs91xGWgolBdmMKchizNBsxhRkUzIsO1iWzeihWZovrQsKExFJOtkZqYzNyOn1Sv7m1jZqDzayc38jO/Y1sHN/A1v3HmLr3kNs23uIBWtrqD5weOAUDslgSGYqmWmpZKWnkJWWSuZhz6lkpqUwvjCH86aPTPpZBRQmIjJopaemMHpoNqOHZsO4rts0tbS9FzJ7QiGzfX8Dh5paaWwJHR9qf95b3/zusaOG5jYam1s50NjCj/6yiuNKhnL+saP4yPRRTBqRO7ADHQA6AC8iEkUbaut4duUOnlm5g9c37QVg0ohcLgiCZfqY/Lg9RqOzuTpRmIhIPNi+7xDPrdzJMyt2sHD9Ltocxg7L5vzpozj7mJEU52WQlpJCeloK6SlGemoKaamh5/TUlAG/Rkdh0onCRETize66Jv66aifPrNzBy2traWpt67VPikFaagppKe3X4di71+GkpvC+Ze2vf/KJ45hVWnhENerUYBGROFc4JIOLZ43j4lnjONDQzOINu6lrbKW5tY2WVqeptY2W1jaaW53mtjaaW5yWtjaaWttobXVa3WlrCz23tvHu6/eWhU6lzskYmDPPFCYiIjGWl5XOWVNHxrqMiOhyUBERiZjCREREIqYwERGRiClMREQkYgoTERGJmMJEREQipjAREZGIKUxERCRig2I6FTOrATYeYfcioLYfy4kHyTamZBsPJN+Ykm08kHxj6mo8E9y9OJzOgyJMImFmleHOTZMokm1MyTYeSL4xJdt4IPnGFOl4tJtLREQipjAREZGIKUx6d2esC4iCZBtTso0Hkm9MyTYeSL4xRTQeHTMREZGIactEREQipjAREZGIDeowMbMsM1tkZm+Y2Uoz+2GHz641s9XB8ps6LP+umVUFn30kNpV3r69jMrNSMztkZsuCxx2xq/5w3Y3HzP7coeYNZrasQ5+EXEfdjSmB19EMM3stqLnSzCo69EnUddTlmBJ4HZ1gZq+a2Ztm9qSZ5Xfo07d15O6D9gEYkBu8TgcWAqcAZwJ/BTKDz0YEz9OAN4BMoAx4B0iN9TgiHFMpsCLWdfd1PJ3a/AL4QaKvox7GlJDrCHgOuCBYPht4KdHXUQ9jStR1tBg4I1h+JfDjI11Hg3rLxEMOBm/Tg4cDXwZudPfGoF110GYO8KC7N7r7eqAKqCCOHMGY4loP4wHAzAy4GPhTsCiR1xHQ5ZjiWg/jcaD9L92hwLbgdSKvo+7GFNd6GM/RwIJg+fPAJ4PXfV5HgzpMAMwsNdidUA087+4LgSnAaWa20Mzmm9msoHkJsLlD9y3BsrjSxzEBlJnZ68Hy02JRc0+6GU+704Cd7r42eJ/I66hd5zFBYq6jrwM/M7PNwM+B7wbNE3kdfZ2uxwSJuY5WABcFTT4NjAte93kdDfowcfdWd58BjAUqzOxYIA0YRmgz8FvAQ8Ffi9bVVwxUreHq45i2A+Pd/UTgG8ADHfebxoNuxtPuUt7/F3wir6N2nceUqOvoy8D17j4OuB64K2ieyOuouzEl6jq6EviqmS0B8oCmoHmf19GgD5N27r4XeAk4n1AKPxpsGi4C2ghNgraF95IbQislbjdzwxlTsBm7K2i/hNC+0SmxqbhnncaDmaUBnwD+3KFZIq+jLseUwOvocuDR4KOHeW83SSKvoy7HlKjryN3fdvfz3P0kQn/AvBM06/M6GtRhYmbFZlYQvM4GzgHeBh4HzgqWTwEyCM2mOQ+4xMwyzawMmAwsGvjKu9fXMQXtU4PlEwmNad3AV961HsZD+2t339KhSyKvI+hiTAm8jrYBZwTNzgLad9sl8jrqckyJuo7MbESwLAX4PtB+Flqf11FalGpPFKOB+4L/E6QAD7n7X8wsA7jbzFYQ2uy73EOnOKw0s4eAVUAL8FV3b41V8d3o05jM7HTgR2bWArQCc919d8yqP1yX4wk+u4ROB6ndPWHXUfDZYWMCEnIdmdle4FfB1lYDcDUk9jrqbkwk7jq6zsy+GrR5FLgHjmwdaToVERGJ2KDezSUiIv1DYSIiIhFTmIiISMQUJiIiEjGFiYiIRExhItILMzvYe6vYS5Q6JTkpTETiTHANg0hCUZiIHAF7774Wy83sMTMbFiz/mpmtCpY/GCw7w967z8XrZpbXxffda2a/NLMXgZ+aWYWZ/SNo/w8zOzpo90Uze9TMnjGztdbhXjsdvqvIQveo+GiU/2cQeZcuWhTphZkddPfcTsuWA9e6+3wz+xGQ7+5fN7NtQJm7N5pZgbvvNbMnCU3//4qZ5QIN7t7S6fvuJTT/2xx3bw0mCax39xYzOwf4srt/0sy+CPwAOBFoBFYDp7r75mA311GEpsL4vrs/H8X/WUTeR1smIn1kZkOBAnefHyy6j9B0GgDLgT+a2ecJTUMB8ArwSzP7WtDvfUHSwcMdpqwYCjwcTH9zMzC9Q7sX3H2fuzcQmu5iQrA8HXgB+LaCRAaawkSkf30UuBU4CVhiZmnufiPwJSAbeM3MpprZ/2vf9dWhb12H1z8GXnT3Y4GPAVkdPmvs8LqV9+bYawGWAHF3G1xJfgoTkT5y933AHnvvBkiXAfODmVfHufuLwLeBAiDXzI5y9zfd/adAJTDV3b/n7jOC+0t0ZSiwNXj9xXBLI3R/iqlmdkNfxyUSCZ01ItK7HDPrOM39Lwnd1+IOM8shNNX4FUAq8IdgN5gBNwfHTH5sZmcS2opYBTwdxm/eRGiW128Afwu30OB4yyXAk2a2391vC7evSCR0AF5ERCKm3VwiIhIxhYmIiERMYSIiIhFTmIiISMQUJiIiEjGFiYiIRExhIiIiEfv/7WPTxiaT1o0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[360:390]\n",
    "x_axis = np.arange(360,390)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "35e24eaa",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[378, 1.0, 0.8150537634408602, 0.8981042654028437, 0.9671880961465089, 1.0, 0.9833204034134988, 0.9407123344081713]\n",
      "false positives 0\n"
     ]
    }
   ],
   "source": [
    "j= 378\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "            f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc1ae201",
   "metadata": {},
   "source": [
    "### GAT Large, wide, huge, small 5, 1%"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 417,
   "id": "78c27902",
   "metadata": {},
   "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>case_id</th>\n",
       "      <th>label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>9000</th>\n",
       "      <td>9001</td>\n",
       "      <td>normal</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "      case_id   label\n",
       "9000     9001  normal"
      ]
     },
     "execution_count": 417,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import the truth label vactor\n",
    "truth_label_path = r'D:\\Final master thesis evaluation\\small-one\\small-0.0-1.csv'\n",
    "anomaly_scores_path = r'D:\\Final master thesis evaluation\\small-one\\small1besthp_test_loss.pt'\n",
    "label_csv = pd.read_csv(truth_label_path)\n",
    "label_csv.drop(\"Unnamed: 0\", axis=1, inplace=True)\n",
    "label_csv[\"case_id\"] = pd.to_numeric(label_csv[\"case_id\"])\n",
    "label_csv.case_id.astype(str).astype(int)\n",
    "label_csv.label.astype(str)\n",
    "#label_csv.dtypes\n",
    "validation_csv = label_csv.loc[label_csv['case_id'] > 9000]\n",
    "validation_csv = validation_csv.loc[label_csv['case_id'] < 12001]\n",
    "test_csv =  label_csv.loc[label_csv['case_id'] > 12000]\n",
    "validation_csv.head(1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 418,
   "id": "fba7ac78",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss_df = pd.read_pickle(anomaly_scores_path)\n",
    "# normalize anomaly scores\n",
    "normalized_loss_df = loss_df.copy()\n",
    "normalized_loss_df['Loss'] = (normalized_loss_df['Loss'] - normalized_loss_df['Loss'].min()) / (normalized_loss_df['Loss'].max() - normalized_loss_df['Loss'].min())\n",
    "# sort the normalized scores\n",
    "sorted_loss = normalized_loss_df.sort_values(by=['Loss'], ascending = False, ignore_index = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 419,
   "id": "96219d22",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "first false positive: 23\n",
      "[22, 1.0, 0.8846153846153846, 0.9387755102040816, 0.9989922741014444, 1.0, 0.9994958830448664, 0.969135696624474]\n"
     ]
    }
   ],
   "source": [
    "loss_df[\"case_id\"] = pd.to_numeric(loss_df[\"case_id\"])\n",
    "result = pd.merge(loss_df, test_csv, on=[\"case_id\"])\n",
    "sorted_result = result.sort_values(by=['Loss'], ascending = False, ignore_index = True)\n",
    "sorted_list = sorted_result['label'].tolist()\n",
    "predictions = []\n",
    "for j in range (len(sorted_list)):\n",
    "    if sorted_list[j] == 'normal':\n",
    "        predictions.append(0)\n",
    "    else:\n",
    "        predictions.append(1)\n",
    "prediction_array = np.array(predictions)\n",
    "best_score = 0\n",
    "first_fp = True\n",
    "for j in range (len(sorted_list)):\n",
    "    current_alarms = prediction_array[0:j+1]\n",
    "    current_normals = prediction_array[j+1:]\n",
    "    positives = j+1\n",
    "    true_positives = np.sum(current_alarms)\n",
    "    false_positives = positives - true_positives\n",
    "    if first_fp and false_positives > 0:\n",
    "        first_fp = False\n",
    "        print('first false positive:', j)\n",
    "    negatives = 3000 - positives\n",
    "    false_negatives = np.sum(current_normals)\n",
    "    true_negatives = negatives - false_negatives\n",
    "    precision = true_positives/(true_positives+false_positives)\n",
    "    recall = true_positives/(true_positives+false_negatives)\n",
    "    f1_score = 2*precision*recall/(precision+recall)\n",
    "    if f1_score > best_score:\n",
    "        best_score = f1_score\n",
    "        precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "        recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "        f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "        macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "        best_cut = [j, precision, recall, f1_score, precision_on_normal, recall_on_normal,\n",
    "                    f1_score_on_normal, macro_f1_score]    \n",
    "print(best_cut)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 420,
   "id": "396afcdd",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss_list = sorted_loss['Loss'].tolist()\n",
    "y_axis = np.array(loss_list)\n",
    "x_axis = np.arange(len(validation_csv))\n",
    "nextplot()\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 421,
   "id": "5b3b783f",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[0:40]\n",
    "x_axis = np.arange(0,40)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 422,
   "id": "2da2e775",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "y2_axis = y_axis[15:25]\n",
    "x_axis = np.arange(15,25)\n",
    "plt.xlabel(\"Loss-rank\")\n",
    "plt.ylabel(\"Normalized Loss\")\n",
    "plt.plot(x_axis, y2_axis)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 423,
   "id": "02c7bd50",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[22, 1.0, 0.8846153846153846, 0.9387755102040816, 0.9745762711864407, 0.969135696624474]\n",
      "false positives 0\n"
     ]
    }
   ],
   "source": [
    "j= 22\n",
    "current_alarms = prediction_array[0:j+1]\n",
    "current_normals = prediction_array[j+1:]\n",
    "positives = j+1\n",
    "true_positives = np.sum(current_alarms)\n",
    "false_positives = positives - true_positives\n",
    "negatives = 3000 - positives\n",
    "false_negatives = np.sum(current_normals)\n",
    "true_negatives = negatives - false_negatives\n",
    "precision = true_positives/(true_positives+false_positives)\n",
    "recall = true_positives/(true_positives+false_negatives)\n",
    "f1_score = 2*precision*recall/(precision+recall)\n",
    "fhalf_score = 5/4*precision*recall/(1/4*precision+recall)\n",
    "best_score = f1_score\n",
    "precision_on_normal = true_negatives/(true_negatives+false_negatives)\n",
    "recall_on_normal = true_negatives/(true_negatives+false_positives)\n",
    "f1_score_on_normal = 2*precision_on_normal*recall_on_normal/(precision_on_normal+recall_on_normal)\n",
    "macro_f1_score = (f1_score + f1_score_on_normal)/2\n",
    "best_cut = [j, precision, recall, f1_score, fhalf_score, macro_f1_score]    \n",
    "print(best_cut)\n",
    "print('false positives', false_positives)"
   ]
  }
 ],
 "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": "3.9.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}