anomalous-token-transfers-ml-py/daily_model_stats.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "83014bac-ac70-4a4c-87f2-450e4faf4343",
   "metadata": {},
   "source": [
    "# Calculate `Anomalous Token Transfers Detection Machine Learning Bot`'s Stats with Forta API\n",
    "\n",
    "\n",
    "* [bot README](https://github.com/forta-network/starter-kits/tree/main/anomalous-token-transfers-ml-py/README.md)\n",
    "* [bot details page](https://explorer.forta.network/bot/0x2e51c6a89c2dccc16a813bb0c3bf3bbfe94414b6a0ea3fc650ad2a59e148f3c8)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "abd1bd77-b2e4-43ec-8c13-dab4dcad088f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from datetime import datetime, timedelta\n",
    "\n",
    "import pandas as pd\n",
    "import requests"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8408f053-b692-40d1-80fe-b452adac986d",
   "metadata": {},
   "source": [
    "# 1. Query previous day's bot alerts with the Forta API"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "id": "a13d1bdd-0d5a-41c0-b6c0-cc97f4e39f0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "forta_api = \"https://api.forta.network/graphql\"\n",
    "headers = {\"content-type\": \"application/json\"}\n",
    "\n",
    "# start and end date needs to be in the format: YYYY-MM-DD\n",
    "START_DATE = str((datetime.today() - timedelta(days=1)).date())\n",
    "END_DATE = START_DATE\n",
    "BOT_ID = \"0x2e51c6a89c2dccc16a813bb0c3bf3bbfe94414b6a0ea3fc650ad2a59e148f3c8\"\n",
    "\n",
    "query = \"\"\"\n",
    "query exampleQuery($input: AlertsInput) {\n",
    "  alerts(input: $input) {\n",
    "    alerts {\n",
    "      findingType\n",
    "      severity\n",
    "      metadata\n",
    "      alertId\n",
    "      hash\n",
    "      source {\n",
    "        block {\n",
    "          number\n",
    "        }\n",
    "        transactionHash\n",
    "      }\n",
    "    }\n",
    "    pageInfo {\n",
    "      hasNextPage\n",
    "      endCursor {\n",
    "        blockNumber\n",
    "        alertId\n",
    "      }\n",
    "    }\n",
    "  }\n",
    "}\n",
    "\"\"\"\n",
    "\n",
    "query_variables = {\n",
    "  \"input\": {\n",
    "    \"first\": 9999,\n",
    "    \"bots\": [BOT_ID],\n",
    "    \"blockDateRange\": {\n",
    "      \"startDate\": START_DATE,\n",
    "      \"endDate\": END_DATE\n",
    "    }\n",
    "  }\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "id": "6bf57c86-43b3-4e5d-b123-e4cd4a394a23",
   "metadata": {},
   "outputs": [],
   "source": [
    "all_alerts = []\n",
    "next_page_exists = True\n",
    "\n",
    "while next_page_exists:\n",
    "    # query Forta API\n",
    "    payload = dict(query=query, variables=query_variables)\n",
    "    response = requests.request(\"POST\", forta_api, json=payload, headers=headers)\n",
    "\n",
    "    # collect alerts\n",
    "    data = response.json()['data']['alerts']\n",
    "    alerts = data['alerts']\n",
    "    all_alerts += alerts\n",
    "\n",
    "    # get next page of alerts if it exists\n",
    "    next_page_exists = data['pageInfo']['hasNextPage']\n",
    "    # endCursor contains alert Id and block number.\n",
    "    # This is needed to get the next page of alerts.\n",
    "    end_cursor = data['pageInfo']['endCursor']\n",
    "    query_variables['input']['after'] = end_cursor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "id": "78b84870-be86-42b6-95b4-2007d581e0a2",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.json_normalize(all_alerts)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "id": "cd6c5e23-548d-4767-a7ba-b67564b33a51",
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(f'{START_DATE}_alerts.csv', index=None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "id": "4111a405-5b3b-4d40-a8ad-80062084ecd9",
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove duplicate alerts for each transaction\n",
    "df = df.drop_duplicates('source.transactionHash')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "c281f00c-ed4a-48f8-acd4-2d265011275e",
   "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>findingType</th>\n",
       "      <th>severity</th>\n",
       "      <th>alertId</th>\n",
       "      <th>hash</th>\n",
       "      <th>metadata.ikolfTransfers</th>\n",
       "      <th>metadata.ikolfValue</th>\n",
       "      <th>metadata.wethTransfers</th>\n",
       "      <th>metadata.wethValue</th>\n",
       "      <th>metadata.accountAgeInMinutes</th>\n",
       "      <th>metadata.anomalyScore</th>\n",
       "      <th>...</th>\n",
       "      <th>metadata.copingTransfers</th>\n",
       "      <th>metadata.copingValue</th>\n",
       "      <th>metadata.asanTransfers</th>\n",
       "      <th>metadata.asanValue</th>\n",
       "      <th>metadata.🐱Transfers</th>\n",
       "      <th>metadata.🐱Value</th>\n",
       "      <th>metadata.cashnipTransfers</th>\n",
       "      <th>metadata.cashnipValue</th>\n",
       "      <th>metadata.ctznTransfers</th>\n",
       "      <th>metadata.ctznValue</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>INFORMATION</td>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>0xf55c12e82132659ac0105facfc72c19f55b84065cfb0dc960f1bf986f331d79c</td>\n",
       "      <td>2</td>\n",
       "      <td>1422.725</td>\n",
       "      <td>2</td>\n",
       "      <td>0.197</td>\n",
       "      <td>812337.0833333334</td>\n",
       "      <td>0.313</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>INFORMATION</td>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>0xed033b70db11cbd05169cb32463b728d4f849415ea07719521d066b6b83b4c3d</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1769047.6833333333</td>\n",
       "      <td>0.314</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>INFORMATION</td>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>0xde8daaafd40b9078393f040877b551d2b57be609f8aa0411a02443426e5871d8</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1</td>\n",
       "      <td>0.507</td>\n",
       "      <td>287494.81666666665</td>\n",
       "      <td>0.305</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>INFORMATION</td>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>0xdd55f9153faf5daab3393aefa0a8e0af8c782b59728272b93ea1d9941c4c67cf</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1769047.6833333333</td>\n",
       "      <td>0.328</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>INFORMATION</td>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>0xdd1baf31b7fe485891811f7c60c22e02af46abaac885b9b24a89496eafb1035f</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>1172160.1333333333</td>\n",
       "      <td>0.317</td>\n",
       "      <td>...</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>5 rows × 8122 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "   findingType severity                    alertId  \\\n",
       "0  INFORMATION     INFO  NORMAL-TOKEN-TRANSFERS-TX   \n",
       "1  INFORMATION     INFO  NORMAL-TOKEN-TRANSFERS-TX   \n",
       "2  INFORMATION     INFO  NORMAL-TOKEN-TRANSFERS-TX   \n",
       "3  INFORMATION     INFO  NORMAL-TOKEN-TRANSFERS-TX   \n",
       "4  INFORMATION     INFO  NORMAL-TOKEN-TRANSFERS-TX   \n",
       "\n",
       "                                                                 hash  \\\n",
       "0  0xf55c12e82132659ac0105facfc72c19f55b84065cfb0dc960f1bf986f331d79c   \n",
       "1  0xed033b70db11cbd05169cb32463b728d4f849415ea07719521d066b6b83b4c3d   \n",
       "2  0xde8daaafd40b9078393f040877b551d2b57be609f8aa0411a02443426e5871d8   \n",
       "3  0xdd55f9153faf5daab3393aefa0a8e0af8c782b59728272b93ea1d9941c4c67cf   \n",
       "4  0xdd1baf31b7fe485891811f7c60c22e02af46abaac885b9b24a89496eafb1035f   \n",
       "\n",
       "  metadata.ikolfTransfers metadata.ikolfValue metadata.wethTransfers  \\\n",
       "0                       2            1422.725                      2   \n",
       "1                     NaN                 NaN                    NaN   \n",
       "2                     NaN                 NaN                      1   \n",
       "3                     NaN                 NaN                    NaN   \n",
       "4                     NaN                 NaN                    NaN   \n",
       "\n",
       "  metadata.wethValue metadata.accountAgeInMinutes  metadata.anomalyScore  ...  \\\n",
       "0              0.197            812337.0833333334                  0.313  ...   \n",
       "1                NaN           1769047.6833333333                  0.314  ...   \n",
       "2              0.507           287494.81666666665                  0.305  ...   \n",
       "3                NaN           1769047.6833333333                  0.328  ...   \n",
       "4                NaN           1172160.1333333333                  0.317  ...   \n",
       "\n",
       "  metadata.copingTransfers metadata.copingValue metadata.asanTransfers  \\\n",
       "0                      NaN                  NaN                    NaN   \n",
       "1                      NaN                  NaN                    NaN   \n",
       "2                      NaN                  NaN                    NaN   \n",
       "3                      NaN                  NaN                    NaN   \n",
       "4                      NaN                  NaN                    NaN   \n",
       "\n",
       "  metadata.asanValue metadata.🐱Transfers metadata.🐱Value  \\\n",
       "0                NaN                 NaN             NaN   \n",
       "1                NaN                 NaN             NaN   \n",
       "2                NaN                 NaN             NaN   \n",
       "3                NaN                 NaN             NaN   \n",
       "4                NaN                 NaN             NaN   \n",
       "\n",
       "  metadata.cashnipTransfers metadata.cashnipValue metadata.ctznTransfers  \\\n",
       "0                       NaN                   NaN                    NaN   \n",
       "1                       NaN                   NaN                    NaN   \n",
       "2                       NaN                   NaN                    NaN   \n",
       "3                       NaN                   NaN                    NaN   \n",
       "4                       NaN                   NaN                    NaN   \n",
       "\n",
       "  metadata.ctznValue  \n",
       "0                NaN  \n",
       "1                NaN  \n",
       "2                NaN  \n",
       "3                NaN  \n",
       "4                NaN  \n",
       "\n",
       "[5 rows x 8122 columns]"
      ]
     },
     "execution_count": 62,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a66daf2c-f844-4fc8-a00b-6d6d3e152018",
   "metadata": {},
   "source": [
    "# 2. Gather Stats"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a80ad498-c00c-44ab-a360-a5baded7fb5a",
   "metadata": {},
   "source": [
    "## Severity Level Breakdown Percentage"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "id": "7fa47671-7663-43ab-ac1d-f5750e1fb76d",
   "metadata": {},
   "outputs": [],
   "source": [
    "TOTAL_ALERTS = len(df)\n",
    "\n",
    "severity_level = df.groupby(['severity', 'alertId']).agg({'hash': 'count'}).reset_index().rename(columns={'hash': 'hash_count'})\n",
    "severity_level['percentage'] = (severity_level['hash_count'] / TOTAL_ALERTS) * 100"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "id": "cdc77153-5eb6-4b63-a80e-8d3e15fdda9d",
   "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>severity</th>\n",
       "      <th>alertId</th>\n",
       "      <th>hash_count</th>\n",
       "      <th>percentage</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>CRITICAL</td>\n",
       "      <td>ANOMALOUS-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>589</td>\n",
       "      <td>0.371365</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>INFO</td>\n",
       "      <td>NORMAL-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>157763</td>\n",
       "      <td>99.469749</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>LOW</td>\n",
       "      <td>INVALID-TOKEN-TRANSFERS-TX</td>\n",
       "      <td>252</td>\n",
       "      <td>0.158886</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   severity                       alertId  hash_count  percentage\n",
       "0  CRITICAL  ANOMALOUS-TOKEN-TRANSFERS-TX         589    0.371365\n",
       "1      INFO     NORMAL-TOKEN-TRANSFERS-TX      157763   99.469749\n",
       "2       LOW    INVALID-TOKEN-TRANSFERS-TX         252    0.158886"
      ]
     },
     "execution_count": 64,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "severity_level"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "edf0310e-d889-4764-9ce9-ca43546e2152",
   "metadata": {},
   "source": [
    "**Anomaly Rate** = `0.3714%`\n",
    "\n",
    "**Model Input Generation Issues** = `0.1589%`"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e17303c-c3ba-41ac-acd9-9b64becbc593",
   "metadata": {},
   "source": [
    "## Alert Time Series by Severity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "671dabed-88a9-4c4b-8720-e19dc0e815bf",
   "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": [
    "import numpy as np\n",
    "import matplotlib.pyplot as plt\n",
    "plt.style.use('seaborn-deep')\n",
    "\n",
    "anomalous_alerts = df[df['severity'] == 'CRITICAL']['source.block.number']\n",
    "model_input_issue_alerts = df[df['severity'] == 'LOW']['source.block.number']\n",
    "normal_tx_alerts = df[df['severity'] == 'INFO']['source.block.number']\n",
    "\n",
    "plt.hist([anomalous_alerts, model_input_issue_alerts, normal_tx_alerts],\n",
    "         label=['ANOMALOUS-TOKEN-TRANSFERS-TX', 'INVALID-TOKEN-TRANSFERS-TX', 'NORMAL-TOKEN-TRANSFERS-TX'],\n",
    "         color=['red', 'orange', 'blue'])\n",
    "plt.legend(loc='upper right')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4b8b5319-3570-4add-8a44-219d6842c13b",
   "metadata": {},
   "source": [
    "## Anomaly Rate Deviation Between Training and Production"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "45b3856b-6656-4603-a8b8-42c85ee68aad",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "There's a -4.040010069770901% deviation of anomaly rate from training to production\n"
     ]
    }
   ],
   "source": [
    "TRAINING_ANOMALY_RATE = 0.387\n",
    "PRODUCTION_ANOMALY_RATE = severity_level[severity_level['alertId'] == 'ANOMALOUS-TOKEN-TRANSFERS-TX']['percentage'].values[0]\n",
    "\n",
    "deviation = ((PRODUCTION_ANOMALY_RATE - TRAINING_ANOMALY_RATE) / TRAINING_ANOMALY_RATE) * 100.0\n",
    "print(f\"There's a {deviation}% deviation of anomaly rate from training to production\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "599035fe-ecbc-4e7d-b0d7-b6eec3db30e9",
   "metadata": {},
   "source": [
    "## Anomaly Score Distribution\n",
    "\n",
    "The above deviation can be reduced by tweaking the model's `ANOMALY_THRESHOLD`. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "id": "83f891f1-4130-4005-a15e-71da3d37220b",
   "metadata": {},
   "outputs": [],
   "source": [
    "df['metadata.anomalyScore'] = df['metadata.anomalyScore'].astype(float)\n",
    "anomalies = df[df['alertId'] == 'ANOMALOUS-TOKEN-TRANSFERS-TX']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "e717b1b6-1bde-4c5d-8fc8-5dda97e5ab30",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<AxesSubplot:>"
      ]
     },
     "execution_count": 68,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "anomalies['metadata.anomalyScore'].hist()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "id": "4a1bb656-f158-48f5-b941-8abc88baea5f",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "An `ANOMALY_THRESHOLD` set to 0.499 can adjust the production anomaly rate to 0.3814531789866586%.\n"
     ]
    }
   ],
   "source": [
    "# increasing the threshold = lower anomaly rate\n",
    "# decreasing the threshold = higher anomaly rate\n",
    "NEW_ANOMALY_THRESHOLD = 0.499\n",
    "new_anomaly_count = (df['metadata.anomalyScore'] >= NEW_ANOMALY_THRESHOLD).value_counts()[1]\n",
    "new_anomaly_rate = (new_anomaly_count / TOTAL_ALERTS) * 100\n",
    "\n",
    "print(f\"An `ANOMALY_THRESHOLD` set to {NEW_ANOMALY_THRESHOLD} can adjust the production anomaly rate to {new_anomaly_rate}%.\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "daf4ae3e-c0a2-4849-956e-69f8f4916e62",
   "metadata": {},
   "source": [
    "If this new anomaly threshold is preferred, don't forget to update it in the detection bot and re-deploy the bot. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0614e1bb-20fe-40d4-9715-dfb40980cb8f",
   "metadata": {},
   "source": [
    "## Top 5 transactions with the highest anomaly score"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "id": "7d4cc9ea-3a4f-47a4-9742-56be66f0ef47",
   "metadata": {},
   "outputs": [],
   "source": [
    "pd.set_option('max_colwidth', None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "id": "bb1c7dac-3e48-4c82-b839-af03c6d83e99",
   "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>source.transactionHash</th>\n",
       "      <th>metadata.anomalyScore</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>21484</th>\n",
       "      <td>0x6b1e11fdbb209b3af30fbc4f2e474566a89bc5cda12948291907640ce028c202</td>\n",
       "      <td>0.727</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17779</th>\n",
       "      <td>0x1bfdbf1ca12fff416b205a4c4be25612b38e70855471c824b9c228fc3fa9351a</td>\n",
       "      <td>0.722</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>56194</th>\n",
       "      <td>0x88c9738e5eefc9912b4bc43fab4cdff1e2da862c799a417968952079d76366e4</td>\n",
       "      <td>0.720</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>98421</th>\n",
       "      <td>0x4025c8b6c6975be1932428866dc3e352446cddaf4c23a864b0607e21a790af6e</td>\n",
       "      <td>0.716</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47548</th>\n",
       "      <td>0x75b83ba93d1331fc01751ff5cf7d595e85a8a906bfc9797928d3af762f543840</td>\n",
       "      <td>0.712</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                   source.transactionHash  \\\n",
       "21484  0x6b1e11fdbb209b3af30fbc4f2e474566a89bc5cda12948291907640ce028c202   \n",
       "17779  0x1bfdbf1ca12fff416b205a4c4be25612b38e70855471c824b9c228fc3fa9351a   \n",
       "56194  0x88c9738e5eefc9912b4bc43fab4cdff1e2da862c799a417968952079d76366e4   \n",
       "98421  0x4025c8b6c6975be1932428866dc3e352446cddaf4c23a864b0607e21a790af6e   \n",
       "47548  0x75b83ba93d1331fc01751ff5cf7d595e85a8a906bfc9797928d3af762f543840   \n",
       "\n",
       "       metadata.anomalyScore  \n",
       "21484                  0.727  \n",
       "17779                  0.722  \n",
       "56194                  0.720  \n",
       "98421                  0.716  \n",
       "47548                  0.712  "
      ]
     },
     "execution_count": 71,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "anomalies.sort_values('metadata.anomalyScore', ascending=False)[['source.transactionHash', 'metadata.anomalyScore']].head(5)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "47cbc129-8981-473a-a832-174141dcd7ff",
   "metadata": {},
   "source": [
    "Transaction `0x6b1e11fdbb209b3af30fbc4f2e474566a89bc5cda12948291907640ce028c202` on Etherscan: [link](https://etherscan.io/tx/0x6b1e11fdbb209b3af30fbc4f2e474566a89bc5cda12948291907640ce028c202)"
   ]
  }
 ],
 "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": "3.8.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}