{
  "id": 549151,
  "title": "CV vs. LB Discrepancy: Is Model Underfitting or Is  CV Method Flawed?",
  "url": "/competitions/child-mind-institute-problematic-internet-use/discussion/549151",
  "author_name": "",
  "post_date": "2024-12-01T01:09:13.852742300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>My model's CV score is 0.377, but the LB score is 0.411. Do you think this is due to an issue with my CV method, or is my model underfitting?</p>",
  "messages": [
    {
      "id": "3059677",
      "postDate": "12/01/2024 01:09:13",
      "content": "<p>My model's CV score is 0.377, but the LB score is 0.411. Do you think this is due to an issue with my CV method, or is my model underfitting?</p>",
      "rawMarkdown": "My model's CV score is 0.377, but the LB score is 0.411. Do you think this is due to an issue with my CV method, or is my model underfitting?",
      "votes": null
    },
    {
      "id": "3060717",
      "postDate": "12/02/2024 03:24:19",
      "content": "<p><a href=\"https://www.kaggle.com/trcnveli\" target=\"_blank\">@trcnveli</a> <br>\nI don't know the details of your model, so I can't say. But I think most of the high scorers on the leaderboard have lower CV scores due to overfitting.<br>\n<a href=\"https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/535523\" target=\"_blank\">This discussion</a> will be helpful.</p>",
      "rawMarkdown": "trcnveli \nI don't know the details of your model, so I can't say. But I think most of the high scorers on the leaderboard have lower CV scores due to overfitting.\n[This discussion](https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/535523) will be helpful.",
      "votes": null
    },
    {
      "id": "3061638",
      "postDate": "12/02/2024 21:36:05",
      "content": "<p>I see CV scores across folds vary a lot which is a bad sign for predictive stability.  I think the essence of the problem is in outliers and that two parents might give very different survey results even though the other health measures of the students are similar, in other words the survey is not capturing something super meaningful…</p>",
      "rawMarkdown": "I see CV scores across folds vary a lot which is a bad sign for predictive stability.  I think the essence of the problem is in outliers and that two parents might give very different survey results even though the other health measures of the students are similar, in other words the survey is not capturing something super meaningful...",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3060717,
      "author_name": "yuogawa",
      "author_url": "",
      "post_date": "12/02/2024 03:24:19",
      "content": "<p><a href=\"https://www.kaggle.com/trcnveli\" target=\"_blank\">@trcnveli</a> <br>\nI don't know the details of your model, so I can't say. But I think most of the high scorers on the leaderboard have lower CV scores due to overfitting.<br>\n<a href=\"https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/535523\" target=\"_blank\">This discussion</a> will be helpful.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3061638,
      "author_name": "lawrencechernin",
      "author_url": "",
      "post_date": "12/02/2024 21:36:05",
      "content": "<p>I see CV scores across folds vary a lot which is a bad sign for predictive stability.  I think the essence of the problem is in outliers and that two parents might give very different survey results even though the other health measures of the students are similar, in other words the survey is not capturing something super meaningful…</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3059677": "My model's CV score is 0.377, but the LB score is 0.411. Do you think this is due to an issue with my CV method, or is my model underfitting?",
    "3060717": "trcnveli \nI don't know the details of your model, so I can't say. But I think most of the high scorers on the leaderboard have lower CV scores due to overfitting.\n[This discussion](https://www.kaggle.com/competitions/child-mind-institute-problematic-internet-use/discussion/535523) will be helpful.",
    "3061638": "I see CV scores across folds vary a lot which is a bad sign for predictive stability.  I think the essence of the problem is in outliers and that two parents might give very different survey results even though the other health measures of the students are similar, in other words the survey is not capturing something super meaningful..."
  },
  "source": "meta"
}