{
  "id": 541199,
  "title": "Regarding date_id on test and forecast data",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/541199",
  "author_name": "",
  "post_date": "2024-10-18T07:06:34.088488100Z",
  "votes": 7,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I want to clarify, will the date_id on the test and forecast data be consistent with the train set and will it always be consistent over time? For example if the last date_id on the train set is 2000, will the test and forecast date_id be always &gt;2000 (2001, 2002, …) ? Or is it possible to have the date_id reset to 0 or will always re-indexed to 0 or some other value? This is may affect models that uses a window or a lookback when evaluating. </p>",
  "messages": [
    {
      "id": "3021070",
      "postDate": "10/18/2024 07:06:34",
      "content": "<p>I want to clarify, will the date_id on the test and forecast data be consistent with the train set and will it always be consistent over time? For example if the last date_id on the train set is 2000, will the test and forecast date_id be always &gt;2000 (2001, 2002, …) ? Or is it possible to have the date_id reset to 0 or will always re-indexed to 0 or some other value? This is may affect models that uses a window or a lookback when evaluating. </p>",
      "rawMarkdown": "I want to clarify, will the date_id on the test and forecast data be consistent with the train set and will it always be consistent over time? For example if the last date_id on the train set is 2000, will the test and forecast date_id be always >2000 (2001, 2002, ...) ? Or is it possible to have the date_id reset to 0 or will always re-indexed to 0 or some other value? This is may affect models that uses a window or a lookback when evaluating.",
      "votes": null
    },
    {
      "id": "3051698",
      "postDate": "11/21/2024 14:59:29",
      "content": "<p>Good question. I was thinking the same. In the test data provided, the <code>date_id</code> is 0 and in the train data the the <code>date_id</code> runs from 0 to 1698. So I'm wondering if the data with <code>date_id</code> = 0 is close to the test data with <code>date_id</code> = 0. </p>",
      "rawMarkdown": "Good question. I was thinking the same. In the test data provided, the `date_id` is 0 and in the train data the the `date_id` runs from 0 to 1698. So I'm wondering if the data with `date_id` = 0 is close to the test data with `date_id` = 0.",
      "votes": null
    },
    {
      "id": "3056490",
      "postDate": "11/27/2024 03:01:07",
      "content": "<p>Same question here, any answer?</p>",
      "rawMarkdown": "Same question here, any answer?",
      "votes": null
    },
    {
      "id": "3056496",
      "postDate": "11/27/2024 03:17:43",
      "content": "<p>I have checked the lag data, and I find that the obs with date_id = 0 in the test set follows the obs with date_id = 1698 in the training set.</p>",
      "rawMarkdown": "I have checked the lag data, and I find that the obs with date_id = 0 in the test set follows the obs with date_id = 1698 in the training set.",
      "votes": null
    },
    {
      "id": "3088115",
      "postDate": "01/04/2025 10:45:57",
      "content": "<p>Considering that submitted notebooks don't show output, how have you checked this? Also, are you saying that the data provided during forecasting follows on the dates from the training set, and yet are labelled <code>date_id=0</code>?</p>",
      "rawMarkdown": "Considering that submitted notebooks don't show output, how have you checked this? Also, are you saying that the data provided during forecasting follows on the dates from the training set, and yet are labelled `date_id=0`?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3051698,
      "author_name": "tbs1980",
      "author_url": "",
      "post_date": "11/21/2024 14:59:29",
      "content": "<p>Good question. I was thinking the same. In the test data provided, the <code>date_id</code> is 0 and in the train data the the <code>date_id</code> runs from 0 to 1698. So I'm wondering if the data with <code>date_id</code> = 0 is close to the test data with <code>date_id</code> = 0. </p>",
      "votes": null,
      "replies": [
        {
          "id": 3056496,
          "author_name": "metropolis40",
          "author_url": "",
          "post_date": "11/27/2024 03:17:43",
          "content": "<p>I have checked the lag data, and I find that the obs with date_id = 0 in the test set follows the obs with date_id = 1698 in the training set.</p>",
          "votes": null,
          "replies": [
            {
              "id": 3088115,
              "author_name": "thijsvanweezel",
              "author_url": "",
              "post_date": "01/04/2025 10:45:57",
              "content": "<p>Considering that submitted notebooks don't show output, how have you checked this? Also, are you saying that the data provided during forecasting follows on the dates from the training set, and yet are labelled <code>date_id=0</code>?</p>",
              "votes": null,
              "replies": []
            }
          ]
        }
      ]
    },
    {
      "id": 3056490,
      "author_name": "kkg715",
      "author_url": "",
      "post_date": "11/27/2024 03:01:07",
      "content": "<p>Same question here, any answer?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3021070": "I want to clarify, will the date_id on the test and forecast data be consistent with the train set and will it always be consistent over time? For example if the last date_id on the train set is 2000, will the test and forecast date_id be always >2000 (2001, 2002, ...) ? Or is it possible to have the date_id reset to 0 or will always re-indexed to 0 or some other value? This is may affect models that uses a window or a lookback when evaluating.",
    "3051698": "Good question. I was thinking the same. In the test data provided, the `date_id` is 0 and in the train data the the `date_id` runs from 0 to 1698. So I'm wondering if the data with `date_id` = 0 is close to the test data with `date_id` = 0.",
    "3056490": "Same question here, any answer?",
    "3056496": "I have checked the lag data, and I find that the obs with date_id = 0 in the test set follows the obs with date_id = 1698 in the training set.",
    "3088115": "Considering that submitted notebooks don't show output, how have you checked this? Also, are you saying that the data provided during forecasting follows on the dates from the training set, and yet are labelled `date_id=0`?"
  },
  "source": "meta"
}