{
  "id": 113679,
  "title": "Avoiding overfitting.",
  "url": "/competitions/ashrae-energy-prediction/discussion/113679",
  "author_name": "",
  "post_date": "2019-10-21T12:22:31.495706900Z",
  "votes": 2,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I made two models.\nOne using month,day from timestamp - CV increased but LB dropped. (Overfitting)\nOther without them - The score was better.</p>\n\n<p>An advice is to force the model to use the weather data instead of dates, to avoid overfitting to the past history since we have only 1years data</p>",
  "messages": [
    {
      "id": "654081",
      "postDate": "10/21/2019 12:22:31",
      "content": "<p>I made two models.\nOne using month,day from timestamp - CV increased but LB dropped. (Overfitting)\nOther without them - The score was better.</p>\n\n<p>An advice is to force the model to use the weather data instead of dates, to avoid overfitting to the past history since we have only 1years data</p>",
      "rawMarkdown": "I made two models.\nOne using month,day from timestamp - CV increased but LB dropped. (Overfitting)\nOther without them - The score was better.\n\nAn advice is to force the model to use the weather data instead of dates, to avoid overfitting to the past history since we have only 1years data",
      "votes": null
    },
    {
      "id": "654082",
      "postDate": "10/21/2019 12:25:09",
      "content": "<p>Re-inventing the wheel and occupying bytes in Kaggle's servers...</p>",
      "rawMarkdown": "Re-inventing the wheel and occupying bytes in Kaggle's servers...",
      "votes": null
    },
    {
      "id": "654086",
      "postDate": "10/21/2019 12:27:07",
      "content": "<p>??</p>",
      "rawMarkdown": "??",
      "votes": null
    },
    {
      "id": "654095",
      "postDate": "10/21/2019 12:39:05",
      "content": "<p>Hahaha this competition is not efficient at all</p>",
      "rawMarkdown": "Hahaha this competition is not efficient at all",
      "votes": null
    },
    {
      "id": "654421",
      "postDate": "10/21/2019 21:09:27",
      "content": "<p>I confirm that removing month helps improving my LB 🤘 </p>",
      "rawMarkdown": "I confirm that removing month helps improving my LB 🤘",
      "votes": null
    },
    {
      "id": "655090",
      "postDate": "10/22/2019 17:17:36",
      "content": "<p><a href=\"/jielu0728\">@jielu0728</a>  :) An upvote will be appreciated!!</p>",
      "rawMarkdown": "jielu0728  :) An upvote will be appreciated!!",
      "votes": null
    },
    {
      "id": "657611",
      "postDate": "10/25/2019 09:42:59",
      "content": "<p>Thanks a lot.But i am curious that 'month' is an important feature from the ouput of lgb features_importance.Why?</p>",
      "rawMarkdown": "Thanks a lot.But i am curious that 'month' is an important feature from the ouput of lgb features_importance.Why?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 654082,
      "author_name": "georsara1",
      "author_url": "",
      "post_date": "10/21/2019 12:25:09",
      "content": "<p>Re-inventing the wheel and occupying bytes in Kaggle's servers...</p>",
      "votes": null,
      "replies": [
        {
          "id": 654086,
          "author_name": "themonologue",
          "author_url": "",
          "post_date": "10/21/2019 12:27:07",
          "content": "<p>??</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 654095,
          "author_name": "gunesevitan",
          "author_url": "",
          "post_date": "10/21/2019 12:39:05",
          "content": "<p>Hahaha this competition is not efficient at all</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 654421,
      "author_name": "jielu0728",
      "author_url": "",
      "post_date": "10/21/2019 21:09:27",
      "content": "<p>I confirm that removing month helps improving my LB 🤘 </p>",
      "votes": null,
      "replies": [
        {
          "id": 655090,
          "author_name": "themonologue",
          "author_url": "",
          "post_date": "10/22/2019 17:17:36",
          "content": "<p><a href=\"/jielu0728\">@jielu0728</a>  :) An upvote will be appreciated!!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 657611,
          "author_name": "dxxhhh",
          "author_url": "",
          "post_date": "10/25/2019 09:42:59",
          "content": "<p>Thanks a lot.But i am curious that 'month' is an important feature from the ouput of lgb features_importance.Why?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "654081": "I made two models.\nOne using month,day from timestamp - CV increased but LB dropped. (Overfitting)\nOther without them - The score was better.\n\nAn advice is to force the model to use the weather data instead of dates, to avoid overfitting to the past history since we have only 1years data",
    "654082": "Re-inventing the wheel and occupying bytes in Kaggle's servers...",
    "654086": "??",
    "654095": "Hahaha this competition is not efficient at all",
    "654421": "I confirm that removing month helps improving my LB 🤘",
    "655090": "jielu0728  :) An upvote will be appreciated!!",
    "657611": "Thanks a lot.But i am curious that 'month' is an important feature from the ouput of lgb features_importance.Why?"
  },
  "source": "meta"
}