{
  "id": 478604,
  "title": "How did you use bureau_a_* data ? ",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/478604",
  "author_name": "",
  "post_date": "2024-02-21T14:05:23.076063900Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi ! <br>\nI wanna use bureau_a_{1, 2}_* data for constructing my ML model (lgb).<br>\nBut bureau_a data is too big to put on my memory (Kaggle Notebook 32GB).<br>\nSo I'd like to ask you how to use bureau_a data. (Like using duckdb or using the part of bureau_a data and so on…)<br>\nPlease comment this topic !!</p>",
  "messages": [
    {
      "id": "2661794",
      "postDate": "02/21/2024 14:05:23",
      "content": "<p>Hi ! <br>\nI wanna use bureau_a_{1, 2}_* data for constructing my ML model (lgb).<br>\nBut bureau_a data is too big to put on my memory (Kaggle Notebook 32GB).<br>\nSo I'd like to ask you how to use bureau_a data. (Like using duckdb or using the part of bureau_a data and so on…)<br>\nPlease comment this topic !!</p>",
      "rawMarkdown": "Hi ! \nI wanna use bureau_a_{1, 2}_* data for constructing my ML model (lgb).\nBut bureau_a data is too big to put on my memory (Kaggle Notebook 32GB).\nSo I'd like to ask you how to use bureau_a data. (Like using duckdb or using the part of bureau_a data and so on...)\nPlease comment this topic !!",
      "votes": null
    },
    {
      "id": "2661798",
      "postDate": "02/21/2024 14:06:44",
      "content": "<p>I would suggest creating aggregation features and then process the tables in batches. </p>",
      "rawMarkdown": "I would suggest creating aggregation features and then process the tables in batches.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2661798,
      "author_name": "jetakow",
      "author_url": "",
      "post_date": "02/21/2024 14:06:44",
      "content": "<p>I would suggest creating aggregation features and then process the tables in batches. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2661794": "Hi ! \nI wanna use bureau_a_{1, 2}_* data for constructing my ML model (lgb).\nBut bureau_a data is too big to put on my memory (Kaggle Notebook 32GB).\nSo I'd like to ask you how to use bureau_a data. (Like using duckdb or using the part of bureau_a data and so on...)\nPlease comment this topic !!",
    "2661798": "I would suggest creating aggregation features and then process the tables in batches."
  },
  "source": "meta"
}