{
  "id": 554166,
  "title": "How to create sequence data for deep learning model",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/554166",
  "author_name": "",
  "post_date": "2024-12-30T20:57:20.087970700Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>When we create sequence data as the input of deep model, the timesteps here, are they created by the same time_id on different days or on the same date but different time_id? I think creating sequence samples by both methods should be fine. Just curious about which format will get better results?</p>",
  "messages": [
    {
      "id": "3084467",
      "postDate": "12/30/2024 20:57:20",
      "content": "<p>When we create sequence data as the input of deep model, the timesteps here, are they created by the same time_id on different days or on the same date but different time_id? I think creating sequence samples by both methods should be fine. Just curious about which format will get better results?</p>",
      "rawMarkdown": "When we create sequence data as the input of deep model, the timesteps here, are they created by the same time_id on different days or on the same date but different time_id? I think creating sequence samples by both methods should be fine. Just curious about which format will get better results?",
      "votes": null
    },
    {
      "id": "3084702",
      "postDate": "12/31/2024 07:15:50",
      "content": "<p>You could also use different time_ids from previous day(s) as they did here (e.g., by sampling 8 from prev day): <a href=\"https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598\" target=\"_blank\">https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598</a></p>",
      "rawMarkdown": "You could also use different time_ids from previous day(s) as they did here (e.g., by sampling 8 from prev day): https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3084702,
      "author_name": "keakohv",
      "author_url": "",
      "post_date": "12/31/2024 07:15:50",
      "content": "<p>You could also use different time_ids from previous day(s) as they did here (e.g., by sampling 8 from prev day): <a href=\"https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598\" target=\"_blank\">https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3084467": "When we create sequence data as the input of deep model, the timesteps here, are they created by the same time_id on different days or on the same date but different time_id? I think creating sequence samples by both methods should be fine. Just curious about which format will get better results?",
    "3084702": "You could also use different time_ids from previous day(s) as they did here (e.g., by sampling 8 from prev day): https://www.kaggle.com/competitions/jane-street-real-time-market-data-forecasting/discussion/553598"
  },
  "source": "meta"
}