{
  "id": 123502,
  "title": "what's the data did private  remove?",
  "url": "/competitions/ashrae-energy-prediction/discussion/123502",
  "author_name": "",
  "post_date": "2019-12-28T07:18:50.566600100Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I think kaggle should just remove leak data and not all data which from  site 0, 1 , 2,  4,15。 but  I think . kaggle remove all data whice site site is from  0,1,2,4,15。 </p>",
  "messages": [
    {
      "id": "704936",
      "postDate": "12/28/2019 07:18:50",
      "content": "<p>I think kaggle should just remove leak data and not all data which from  site 0, 1 , 2,  4,15。 but  I think . kaggle remove all data whice site site is from  0,1,2,4,15。 </p>",
      "rawMarkdown": "I think kaggle should just remove leak data and not all data which from  site 0, 1 , 2,  4,15。 but  I think . kaggle remove all data whice site site is from  0,1,2,4,15。",
      "votes": null
    },
    {
      "id": "704938",
      "postDate": "12/28/2019 07:25:27",
      "content": "<p>I think this is  the reason make our  private became lower, because  we use leak data as validataion.</p>",
      "rawMarkdown": "I think this is  the reason make our  private became lower, because  we use leak data as validataion.",
      "votes": null
    },
    {
      "id": "704941",
      "postDate": "12/28/2019 07:27:41",
      "content": "<p><a href=\"/sohier\">@sohier</a> </p>",
      "rawMarkdown": "sohier",
      "votes": null
    },
    {
      "id": "704948",
      "postDate": "12/28/2019 07:44:28",
      "content": "<p>once a site is compromised, it means any predictions for that site is untrustworthy. even if not all the buildings or meters for a particular site is found, there can still be intra-site_id leakage (some buildings are correlated). If Kaggle admins didn't remove all available sites posted on external data page, then the final private LB wouldn't be objective as possible and thus render the results less meaningful.</p>",
      "rawMarkdown": "once a site is compromised, it means any predictions for that site is untrustworthy. even if not all the buildings or meters for a particular site is found, there can still be intra-site_id leakage (some buildings are correlated). If Kaggle admins didn't remove all available sites posted on external data page, then the final private LB wouldn't be objective as possible and thus render the results less meaningful.",
      "votes": null
    },
    {
      "id": "709608",
      "postDate": "01/03/2020 18:04:10",
      "content": "<p>Each site was either completely included or completely excluded from the private LB for the reasons <a href=\"/teeyee314\">@teeyee314</a> noted.</p>",
      "rawMarkdown": "Each site was either completely included or completely excluded from the private LB for the reasons @teeyee314 noted.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 704938,
      "author_name": "bestpredict",
      "author_url": "",
      "post_date": "12/28/2019 07:25:27",
      "content": "<p>I think this is  the reason make our  private became lower, because  we use leak data as validataion.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 704941,
      "author_name": "bestpredict",
      "author_url": "",
      "post_date": "12/28/2019 07:27:41",
      "content": "<p><a href=\"/sohier\">@sohier</a> </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 704948,
      "author_name": "teeyee314",
      "author_url": "",
      "post_date": "12/28/2019 07:44:28",
      "content": "<p>once a site is compromised, it means any predictions for that site is untrustworthy. even if not all the buildings or meters for a particular site is found, there can still be intra-site_id leakage (some buildings are correlated). If Kaggle admins didn't remove all available sites posted on external data page, then the final private LB wouldn't be objective as possible and thus render the results less meaningful.</p>",
      "votes": null,
      "replies": [
        {
          "id": 709608,
          "author_name": "sohier",
          "author_url": "",
          "post_date": "01/03/2020 18:04:10",
          "content": "<p>Each site was either completely included or completely excluded from the private LB for the reasons <a href=\"/teeyee314\">@teeyee314</a> noted.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "704936": "I think kaggle should just remove leak data and not all data which from  site 0, 1 , 2,  4,15。 but  I think . kaggle remove all data whice site site is from  0,1,2,4,15。",
    "704938": "I think this is  the reason make our  private became lower, because  we use leak data as validataion.",
    "704941": "sohier",
    "704948": "once a site is compromised, it means any predictions for that site is untrustworthy. even if not all the buildings or meters for a particular site is found, there can still be intra-site_id leakage (some buildings are correlated). If Kaggle admins didn't remove all available sites posted on external data page, then the final private LB wouldn't be objective as possible and thus render the results less meaningful.",
    "709608": "Each site was either completely included or completely excluded from the private LB for the reasons @teeyee314 noted."
  },
  "source": "meta"
}