{
  "id": 580633,
  "title": "Train and Test Distribution",
  "url": "/competitions/drw-crypto-market-prediction/discussion/580633",
  "author_name": "",
  "post_date": "2025-05-25T15:39:34.767649300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Based on some EDA, I think the train and test data are from different distributions, could someone please guide me as to how this should be dealt with or maybe if I'm going wrong somewhere in my approach.</p>",
  "messages": [
    {
      "id": "3209331",
      "postDate": "05/25/2025 15:39:34",
      "content": "<p>Based on some EDA, I think the train and test data are from different distributions, could someone please guide me as to how this should be dealt with or maybe if I'm going wrong somewhere in my approach.</p>",
      "rawMarkdown": "Based on some EDA, I think the train and test data are from different distributions, could someone please guide me as to how this should be dealt with or maybe if I'm going wrong somewhere in my approach.",
      "votes": null
    },
    {
      "id": "3209510",
      "postDate": "05/26/2025 00:33:46",
      "content": "<p>It's normal,because it's time series.For example,one feature in train set is temperature,the range of train set is March to July,the range of test set is August to October.</p>",
      "rawMarkdown": "It's normal,because it's time series.For example,one feature in train set is temperature,the range of train set is March to July,the range of test set is August to October.",
      "votes": null
    },
    {
      "id": "3209695",
      "postDate": "05/26/2025 07:22:56",
      "content": "<p>financial time series almost always has domain-shift. try not to overfit.</p>",
      "rawMarkdown": "financial time series almost always has domain-shift. try not to overfit.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3209510,
      "author_name": "yunsuxiaozi",
      "author_url": "",
      "post_date": "05/26/2025 00:33:46",
      "content": "<p>It's normal,because it's time series.For example,one feature in train set is temperature,the range of train set is March to July,the range of test set is August to October.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3209695,
      "author_name": "henrysun",
      "author_url": "",
      "post_date": "05/26/2025 07:22:56",
      "content": "<p>financial time series almost always has domain-shift. try not to overfit.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3209331": "Based on some EDA, I think the train and test data are from different distributions, could someone please guide me as to how this should be dealt with or maybe if I'm going wrong somewhere in my approach.",
    "3209510": "It's normal,because it's time series.For example,one feature in train set is temperature,the range of train set is March to July,the range of test set is August to October.",
    "3209695": "financial time series almost always has domain-shift. try not to overfit."
  },
  "source": "meta"
}