{
  "id": 554284,
  "title": "Adjusting date_id in test data to align with train dataset",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/554284",
  "author_name": "Calvin Hui",
  "post_date": "2024-12-31T16:22:03.557000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi all,</p>\n<p>I used <code>date_id</code> and <code>time_id</code> as features when I was training my LightGBM model. However, I noticed that in the <code>test.parquet</code> file, the <code>date_id</code> starts from 0. I think I need to make some adjustments to align the <code>date_id</code> values in the test dataset with those in the train dataset.</p>\n<p>However, I'm not sure whether the test data continues directly after the train data. If this is the case, I can simply add 1699 (as the last <code>date_id</code> value provided in the <code>train.parquet</code> is 1698) to every <code>date_id</code> value in the test dataset.</p>\n<p>I’d appreciate any insights or confirmation on whether this assumption holds true.</p>",
  "messages": [
    {
      "id": 3085056,
      "postDate": "2024-12-31T16:22:03.557Z",
      "content": "<p>Hi all,</p>\n<p>I used <code>date_id</code> and <code>time_id</code> as features when I was training my LightGBM model. However, I noticed that in the <code>test.parquet</code> file, the <code>date_id</code> starts from 0. I think I need to make some adjustments to align the <code>date_id</code> values in the test dataset with those in the train dataset.</p>\n<p>However, I'm not sure whether the test data continues directly after the train data. If this is the case, I can simply add 1699 (as the last <code>date_id</code> value provided in the <code>train.parquet</code> is 1698) to every <code>date_id</code> value in the test dataset.</p>\n<p>I’d appreciate any insights or confirmation on whether this assumption holds true.</p>",
      "rawMarkdown": "Hi all,\n\nI used `date_id` and `time_id` as features when I was training my LightGBM model. However, I noticed that in the `test.parquet` file, the `date_id` starts from 0. I think I need to make some adjustments to align the `date_id` values in the test dataset with those in the train dataset.\n\nHowever, I'm not sure whether the test data continues directly after the train data. If this is the case, I can simply add 1699 (as the last `date_id` value provided in the `train.parquet` is 1698) to every `date_id` value in the test dataset.\n\nI’d appreciate any insights or confirmation on whether this assumption holds true."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "3085056": "Hi all,\n\nI used `date_id` and `time_id` as features when I was training my LightGBM model. However, I noticed that in the `test.parquet` file, the `date_id` starts from 0. I think I need to make some adjustments to align the `date_id` values in the test dataset with those in the train dataset.\n\nHowever, I'm not sure whether the test data continues directly after the train data. If this is the case, I can simply add 1699 (as the last `date_id` value provided in the `train.parquet` is 1698) to every `date_id` value in the test dataset.\n\nI’d appreciate any insights or confirmation on whether this assumption holds true."
  }
}