{
  "id": 545491,
  "title": "Overfitting issues",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/545491",
  "author_name": "",
  "post_date": "2024-11-10T15:14:40.575810700Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I'm struggling with my model overfitting to the training data, which leads to lower performance when it comes to the actual evaluation data set. How are you solving the issue to ensure the models generalizes better for unseen data sets?</p>",
  "messages": [
    {
      "id": "3041604",
      "postDate": "11/10/2024 15:14:40",
      "content": "<p>I'm struggling with my model overfitting to the training data, which leads to lower performance when it comes to the actual evaluation data set. How are you solving the issue to ensure the models generalizes better for unseen data sets?</p>",
      "rawMarkdown": "I'm struggling with my model overfitting to the training data, which leads to lower performance when it comes to the actual evaluation data set. How are you solving the issue to ensure the models generalizes better for unseen data sets?",
      "votes": null
    },
    {
      "id": "3041640",
      "postDate": "11/10/2024 16:36:03",
      "content": "<p>I am also experiencing the same issue :(</p>",
      "rawMarkdown": "I am also experiencing the same issue :(",
      "votes": null
    },
    {
      "id": "3041672",
      "postDate": "11/10/2024 17:16:35",
      "content": "<p>Wait until you see the private LB score <br>\nIf you're using online learning then I can only wonder how you was able to do it <br>\nIf not then I believe you should count for six month gap to get a result that represents what the model will face during private LB <br>\nThe 1 min limit just complicated everything </p>",
      "rawMarkdown": "Wait until you see the private LB score \nIf you're using online learning then I can only wonder how you was able to do it \nIf not then I believe you should count for six month gap to get a result that represents what the model will face during private LB \nThe 1 min limit just complicated everything",
      "votes": null
    },
    {
      "id": "3041815",
      "postDate": "11/10/2024 22:37:09",
      "content": "<p>Agreed <a href=\"https://www.kaggle.com/aymanallawi\" target=\"_blank\">@aymanallawi</a> <br>\nThe duration of the forecast period is far too long. One needs to wait for 6 months after the final model choice to get a medal here. This seems to be a put off.<br>\nAlso the 1 minute rule has complicated a lot for most of us, but we still have a long road ahead for a good model design. A few top solutions have seemingly churned out some magic from the data as on date at least!!</p>",
      "rawMarkdown": "Agreed @aymanallawi \nThe duration of the forecast period is far too long. One needs to wait for 6 months after the final model choice to get a medal here. This seems to be a put off.\nAlso the 1 minute rule has complicated a lot for most of us, but we still have a long road ahead for a good model design. A few top solutions have seemingly churned out some magic from the data as on date at least!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3041640,
      "author_name": "ebtrying",
      "author_url": "",
      "post_date": "11/10/2024 16:36:03",
      "content": "<p>I am also experiencing the same issue :(</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3041672,
      "author_name": "aymanallawi",
      "author_url": "",
      "post_date": "11/10/2024 17:16:35",
      "content": "<p>Wait until you see the private LB score <br>\nIf you're using online learning then I can only wonder how you was able to do it <br>\nIf not then I believe you should count for six month gap to get a result that represents what the model will face during private LB <br>\nThe 1 min limit just complicated everything </p>",
      "votes": null,
      "replies": [
        {
          "id": 3041815,
          "author_name": "ravi20076",
          "author_url": "",
          "post_date": "11/10/2024 22:37:09",
          "content": "<p>Agreed <a href=\"https://www.kaggle.com/aymanallawi\" target=\"_blank\">@aymanallawi</a> <br>\nThe duration of the forecast period is far too long. One needs to wait for 6 months after the final model choice to get a medal here. This seems to be a put off.<br>\nAlso the 1 minute rule has complicated a lot for most of us, but we still have a long road ahead for a good model design. A few top solutions have seemingly churned out some magic from the data as on date at least!!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3041604": "I'm struggling with my model overfitting to the training data, which leads to lower performance when it comes to the actual evaluation data set. How are you solving the issue to ensure the models generalizes better for unseen data sets?",
    "3041640": "I am also experiencing the same issue :(",
    "3041672": "Wait until you see the private LB score \nIf you're using online learning then I can only wonder how you was able to do it \nIf not then I believe you should count for six month gap to get a result that represents what the model will face during private LB \nThe 1 min limit just complicated everything",
    "3041815": "Agreed @aymanallawi \nThe duration of the forecast period is far too long. One needs to wait for 6 months after the final model choice to get a medal here. This seems to be a put off.\nAlso the 1 minute rule has complicated a lot for most of us, but we still have a long road ahead for a good model design. A few top solutions have seemingly churned out some magic from the data as on date at least!!"
  },
  "source": "meta"
}