{
  "id": 541202,
  "title": "How to model the temporal dependency?",
  "url": "/competitions/jane-street-real-time-market-data-forecasting/discussion/541202",
  "author_name": "",
  "post_date": "2024-10-18T07:09:19.641209600Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I find myself puzzled by the notion that time series data might be useless. It appears that when we attempt to make predictions, we're only granted access to the feature data for each specific time point (akin to mock test data) and the responder data with a one-time lag. Consequently, it feels as though we're unable to effectively harness the temporal dependencies within our model. Am I misinterpreting this situation?</p>",
  "messages": [
    {
      "id": "3021076",
      "postDate": "10/18/2024 07:09:19",
      "content": "<p>I find myself puzzled by the notion that time series data might be useless. It appears that when we attempt to make predictions, we're only granted access to the feature data for each specific time point (akin to mock test data) and the responder data with a one-time lag. Consequently, it feels as though we're unable to effectively harness the temporal dependencies within our model. Am I misinterpreting this situation?</p>",
      "rawMarkdown": "I find myself puzzled by the notion that time series data might be useless. It appears that when we attempt to make predictions, we're only granted access to the feature data for each specific time point (akin to mock test data) and the responder data with a one-time lag. Consequently, it feels as though we're unable to effectively harness the temporal dependencies within our model. Am I misinterpreting this situation?",
      "votes": null
    },
    {
      "id": "3021108",
      "postDate": "10/18/2024 07:40:50",
      "content": "<p>you can make your own historical cache to build the time-series in the submission.</p>",
      "rawMarkdown": "you can make your own historical cache to build the time-series in the submission.",
      "votes": null
    },
    {
      "id": "3021223",
      "postDate": "10/18/2024 09:23:48",
      "content": "<p>Great solution, it addressed my question perfectly! Thank you <a href=\"https://www.kaggle.com/shiyili\" target=\"_blank\">@shiyili</a> </p>",
      "rawMarkdown": "Great solution, it addressed my question perfectly! Thank you @shiyili",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3021108,
      "author_name": "shiyili",
      "author_url": "",
      "post_date": "10/18/2024 07:40:50",
      "content": "<p>you can make your own historical cache to build the time-series in the submission.</p>",
      "votes": null,
      "replies": [
        {
          "id": 3021223,
          "author_name": "yicheung1",
          "author_url": "",
          "post_date": "10/18/2024 09:23:48",
          "content": "<p>Great solution, it addressed my question perfectly! Thank you <a href=\"https://www.kaggle.com/shiyili\" target=\"_blank\">@shiyili</a> </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "3021076": "I find myself puzzled by the notion that time series data might be useless. It appears that when we attempt to make predictions, we're only granted access to the feature data for each specific time point (akin to mock test data) and the responder data with a one-time lag. Consequently, it feels as though we're unable to effectively harness the temporal dependencies within our model. Am I misinterpreting this situation?",
    "3021108": "you can make your own historical cache to build the time-series in the submission.",
    "3021223": "Great solution, it addressed my question perfectly! Thank you @shiyili"
  },
  "source": "meta"
}