{
  "id": 321351,
  "title": "Reporting anomaly annotation issues",
  "url": "/competitions/energy-anomaly-detection/discussion/321351",
  "author_name": "",
  "post_date": "2022-04-26T12:19:24.077506200Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hello team,</p>\n<p>As you all know, it is extremely difficult to collect energy meter readings with true anomalies from operational buildings. Hence, we have annotated the already collected meter readings based on our own expertise and energy domain knowledge. Despite following careful annotation process, there are still chances of error in our anomaly labels.</p>\n<p>I opened this thread to report such incorrectly annotated anomaly labels. You may reply to this thread with the <code>building_id</code> and <code>timestamp</code> of wrong anomaly instances to help us fix the issue. </p>\n<p>Thank you, and happing coding!</p>\n<p><a href=\"https://github.com/samy101/lead-dataset\" target=\"_blank\">The LEAD project team</a></p>",
  "messages": [
    {
      "id": "1768570",
      "postDate": "04/26/2022 12:19:24",
      "content": "<p>Hello team,</p>\n<p>As you all know, it is extremely difficult to collect energy meter readings with true anomalies from operational buildings. Hence, we have annotated the already collected meter readings based on our own expertise and energy domain knowledge. Despite following careful annotation process, there are still chances of error in our anomaly labels.</p>\n<p>I opened this thread to report such incorrectly annotated anomaly labels. You may reply to this thread with the <code>building_id</code> and <code>timestamp</code> of wrong anomaly instances to help us fix the issue. </p>\n<p>Thank you, and happing coding!</p>\n<p><a href=\"https://github.com/samy101/lead-dataset\" target=\"_blank\">The LEAD project team</a></p>",
      "rawMarkdown": "Hello team,\n\nAs you all know, it is extremely difficult to collect energy meter readings with true anomalies from operational buildings. Hence, we have annotated the already collected meter readings based on our own expertise and energy domain knowledge. Despite following careful annotation process, there are still chances of error in our anomaly labels.\n\nI opened this thread to report such incorrectly annotated anomaly labels. You may reply to this thread with the `building_id` and `timestamp` of wrong anomaly instances to help us fix the issue. \n\nThank you, and happing coding!\n\n[The LEAD project team](https://github.com/samy101/lead-dataset)",
      "votes": null
    },
    {
      "id": "1792263",
      "postDate": "05/16/2022 19:20:23",
      "content": "<p>What are you basing the anomaly instances on?  If you look at the below image, you can see that Monday from 12 noon - 2:00 PM is flagged as an anomalies, but none of the other days in that week are flagged.</p>\n<p>I am currently at work and it is hard to upload the file to a 3rd party to share. I will upload image when I get home.</p>\n<p>[<img src=\"https://imgur.com/1x33AWx.png\">]</p>",
      "rawMarkdown": "What are you basing the anomaly instances on?  If you look at the below image, you can see that Monday from 12 noon - 2:00 PM is flagged as an anomalies, but none of the other days in that week are flagged.\n\nI am currently at work and it is hard to upload the file to a 3rd party to share. I will upload image when I get home.\n\n\n\n[<img src=\"https://imgur.com/1x33AWx.png\" width=\"3000px\">]",
      "votes": null
    },
    {
      "id": "1793218",
      "postDate": "05/17/2022 17:12:39",
      "content": "<p><a href=\"https://www.kaggle.com/stephenstrickland\" target=\"_blank\">@stephenstrickland</a>, I quickly checked those readings (see the below image) on Jan 4, 12-2 pm. I think they were tagged as anomalies because there was a lower energy use (a short dip) compared to the neighboring days. As we have mentioned in this post, there are rare chances of errors in our annotated process. Moreover, it is not uncommon to have noisy or outliers in datasets being used in Kaggle competitions!</p>\n<p><img src=\"https://i.imgur.com/JEASxyX.png\" alt=\"\"></p>",
      "rawMarkdown": "stephenstrickland, I quickly checked those readings (see the below image) on Jan 4, 12-2 pm. I think they were tagged as anomalies because there was a lower energy use (a short dip) compared to the neighboring days. As we have mentioned in this post, there are rare chances of errors in our annotated process. Moreover, it is not uncommon to have noisy or outliers in datasets being used in Kaggle competitions!\n\n![](https://i.imgur.com/JEASxyX.png)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1792263,
      "author_name": "stephenstrickland",
      "author_url": "",
      "post_date": "05/16/2022 19:20:23",
      "content": "<p>What are you basing the anomaly instances on?  If you look at the below image, you can see that Monday from 12 noon - 2:00 PM is flagged as an anomalies, but none of the other days in that week are flagged.</p>\n<p>I am currently at work and it is hard to upload the file to a 3rd party to share. I will upload image when I get home.</p>\n<p>[<img src=\"https://imgur.com/1x33AWx.png\">]</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1793218,
      "author_name": "samy101",
      "author_url": "",
      "post_date": "05/17/2022 17:12:39",
      "content": "<p><a href=\"https://www.kaggle.com/stephenstrickland\" target=\"_blank\">@stephenstrickland</a>, I quickly checked those readings (see the below image) on Jan 4, 12-2 pm. I think they were tagged as anomalies because there was a lower energy use (a short dip) compared to the neighboring days. As we have mentioned in this post, there are rare chances of errors in our annotated process. Moreover, it is not uncommon to have noisy or outliers in datasets being used in Kaggle competitions!</p>\n<p><img src=\"https://i.imgur.com/JEASxyX.png\" alt=\"\"></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1768570": "Hello team,\n\nAs you all know, it is extremely difficult to collect energy meter readings with true anomalies from operational buildings. Hence, we have annotated the already collected meter readings based on our own expertise and energy domain knowledge. Despite following careful annotation process, there are still chances of error in our anomaly labels.\n\nI opened this thread to report such incorrectly annotated anomaly labels. You may reply to this thread with the `building_id` and `timestamp` of wrong anomaly instances to help us fix the issue. \n\nThank you, and happing coding!\n\n[The LEAD project team](https://github.com/samy101/lead-dataset)",
    "1792263": "What are you basing the anomaly instances on?  If you look at the below image, you can see that Monday from 12 noon - 2:00 PM is flagged as an anomalies, but none of the other days in that week are flagged.\n\nI am currently at work and it is hard to upload the file to a 3rd party to share. I will upload image when I get home.\n\n\n\n[<img src=\"https://imgur.com/1x33AWx.png\" width=\"3000px\">]",
    "1793218": "stephenstrickland, I quickly checked those readings (see the below image) on Jan 4, 12-2 pm. I think they were tagged as anomalies because there was a lower energy use (a short dip) compared to the neighboring days. As we have mentioned in this post, there are rare chances of errors in our annotated process. Moreover, it is not uncommon to have noisy or outliers in datasets being used in Kaggle competitions!\n\n![](https://i.imgur.com/JEASxyX.png)"
  },
  "source": "meta"
}