{
  "id": 338634,
  "title": "Dropping features affect on cv and LB?",
  "url": "/competitions/amex-default-prediction/discussion/338634",
  "author_name": "",
  "post_date": "2022-07-21T08:49:33.935727100Z",
  "votes": 6,
  "comment_count": 6,
  "views": 0,
  "content": "<p>I have a simple query please respond,<br>\nI dropped some features , it improved my cv , but decreased my lb score..what went wrong?</p>\n<p>After dropping cv increased by 0.001<br>\nbut lb decreased by 0.001</p>\n<p>Single model XGB seed 42 5 Folds<br>\nC.V. - 0.797<br>\nbut LB - 0.795 </p>\n<p>Samething happend with my catboost model .</p>",
  "messages": [
    {
      "id": "1864703",
      "postDate": "07/21/2022 08:49:33",
      "content": "<p>I have a simple query please respond,<br>\nI dropped some features , it improved my cv , but decreased my lb score..what went wrong?</p>\n<p>After dropping cv increased by 0.001<br>\nbut lb decreased by 0.001</p>\n<p>Single model XGB seed 42 5 Folds<br>\nC.V. - 0.797<br>\nbut LB - 0.795 </p>\n<p>Samething happend with my catboost model .</p>",
      "rawMarkdown": "I have a simple query please respond,\nI dropped some features , it improved my cv , but decreased my lb score..what went wrong?\n\nAfter dropping cv increased by 0.001\nbut lb decreased by 0.001\n\nSingle model XGB seed 42 5 Folds\nC.V. - 0.797\nbut LB - 0.795 \n\n\nSamething happend with my catboost model .",
      "votes": null
    },
    {
      "id": "1864818",
      "postDate": "07/21/2022 10:18:36",
      "content": "<ul>\n<li>CV overfit</li>\n<li>Test Deletion of important features in the data</li>\n</ul>",
      "rawMarkdown": "CV overfit\n- Test Deletion of important features in the data",
      "votes": null
    },
    {
      "id": "1864828",
      "postDate": "07/21/2022 10:24:26",
      "content": "<p><a href=\"https://www.kaggle.com/zakopur0\" target=\"_blank\">@zakopur0</a> ohk Thanks</p>",
      "rawMarkdown": "zakopur0 ohk Thanks",
      "votes": null
    },
    {
      "id": "1868415",
      "postDate": "07/24/2022 00:45:30",
      "content": "<p>It sounds like you may be overfitting to your validation data if your CV score improved but your LB score decreased..</p>",
      "rawMarkdown": "It sounds like you may be overfitting to your validation data if your CV score improved but your LB score decreased..",
      "votes": null
    },
    {
      "id": "1868508",
      "postDate": "07/24/2022 02:54:40",
      "content": "<p><a href=\"https://www.kaggle.com/thedevastator\" target=\"_blank\">@thedevastator</a> ohk</p>",
      "rawMarkdown": "thedevastator ohk",
      "votes": null
    },
    {
      "id": "1870378",
      "postDate": "07/25/2022 14:05:13",
      "content": "<p><a href=\"https://www.kaggle.com/kushal1506\" target=\"_blank\">@kushal1506</a> It could be also the opposite, that you overfitted the public LB and now after dropping some features the LB decreased but the CV increased. It is suggested to follow the CV because you are using all the available data to train your model and to validate it on each fold accordingly.</p>",
      "rawMarkdown": "kushal1506 It could be also the opposite, that you overfitted the public LB and now after dropping some features the LB decreased but the CV increased. It is suggested to follow the CV because you are using all the available data to train your model and to validate it on each fold accordingly.",
      "votes": null
    },
    {
      "id": "1872095",
      "postDate": "07/26/2022 17:14:51",
      "content": "<p><a href=\"https://www.kaggle.com/aliabdin1\" target=\"_blank\">@aliabdin1</a> Thank U </p>",
      "rawMarkdown": "aliabdin1 Thank U",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1864818,
      "author_name": "zakopur0",
      "author_url": "",
      "post_date": "07/21/2022 10:18:36",
      "content": "<ul>\n<li>CV overfit</li>\n<li>Test Deletion of important features in the data</li>\n</ul>",
      "votes": null,
      "replies": [
        {
          "id": 1864828,
          "author_name": "kushal1506",
          "author_url": "",
          "post_date": "07/21/2022 10:24:26",
          "content": "<p><a href=\"https://www.kaggle.com/zakopur0\" target=\"_blank\">@zakopur0</a> ohk Thanks</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1868415,
      "author_name": "thedevastator",
      "author_url": "",
      "post_date": "07/24/2022 00:45:30",
      "content": "<p>It sounds like you may be overfitting to your validation data if your CV score improved but your LB score decreased..</p>",
      "votes": null,
      "replies": [
        {
          "id": 1868508,
          "author_name": "kushal1506",
          "author_url": "",
          "post_date": "07/24/2022 02:54:40",
          "content": "<p><a href=\"https://www.kaggle.com/thedevastator\" target=\"_blank\">@thedevastator</a> ohk</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1870378,
          "author_name": "aliabdin1",
          "author_url": "",
          "post_date": "07/25/2022 14:05:13",
          "content": "<p><a href=\"https://www.kaggle.com/kushal1506\" target=\"_blank\">@kushal1506</a> It could be also the opposite, that you overfitted the public LB and now after dropping some features the LB decreased but the CV increased. It is suggested to follow the CV because you are using all the available data to train your model and to validate it on each fold accordingly.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1872095,
          "author_name": "kushal1506",
          "author_url": "",
          "post_date": "07/26/2022 17:14:51",
          "content": "<p><a href=\"https://www.kaggle.com/aliabdin1\" target=\"_blank\">@aliabdin1</a> Thank U </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1864703": "I have a simple query please respond,\nI dropped some features , it improved my cv , but decreased my lb score..what went wrong?\n\nAfter dropping cv increased by 0.001\nbut lb decreased by 0.001\n\nSingle model XGB seed 42 5 Folds\nC.V. - 0.797\nbut LB - 0.795 \n\n\nSamething happend with my catboost model .",
    "1864818": "CV overfit\n- Test Deletion of important features in the data",
    "1864828": "zakopur0 ohk Thanks",
    "1868415": "It sounds like you may be overfitting to your validation data if your CV score improved but your LB score decreased..",
    "1868508": "thedevastator ohk",
    "1870378": "kushal1506 It could be also the opposite, that you overfitted the public LB and now after dropping some features the LB decreased but the CV increased. It is suggested to follow the CV because you are using all the available data to train your model and to validate it on each fold accordingly.",
    "1872095": "aliabdin1 Thank U"
  },
  "source": "meta"
}