{
  "id": 476763,
  "title": "What is the meaning of num_group1 and num_group2 in depth1 and 2 dataset?",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/476763",
  "author_name": "",
  "post_date": "2024-02-13T12:42:46.521175800Z",
  "votes": 16,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I can't understand what is the meaning for num_group1 and num_group2.🫠<br>\nAnyone can help me, thanks a lot.</p>",
  "messages": [
    {
      "id": "2650392",
      "postDate": "02/13/2024 12:42:46",
      "content": "<p>I can't understand what is the meaning for num_group1 and num_group2.🫠<br>\nAnyone can help me, thanks a lot.</p>",
      "rawMarkdown": "I can't understand what is the meaning for num_group1 and num_group2.🫠\nAnyone can help me, thanks a lot.",
      "votes": null
    },
    {
      "id": "2650538",
      "postDate": "02/13/2024 14:09:18",
      "content": "<p>Hi Fang,</p>\n<p>let me try to answer by example: goal of this competition is to predict risk quality of clients (or to be precise applications of clients), so client and his/her application is entity (case_id) for which your model predict probability of default (credit score).</p>\n<ul>\n<li><p>depth=0 (what we call static attributes) are attributes that are aggregated on case_id level. Example can be age of client or gender, 1 records per 1 case_id.</p></li>\n<li><p>depth=1, those are attributes where we have several records per client/application. Example can be previous applications or loans in credit bureau register, each client can have from 0 to n records. Therefore, for one case_id there might be several records, and to index them we use num_group1</p></li>\n<li><p>depth=2: For some attributes with depth=1 we have more detailed information, for example for previous applications we have data about instalments like date of payments or days past due of each payments. It means for each previous application, you can have 0…n records about instalments/payments. And as an index is used num_group2</p></li>\n</ul>\n<p>To sum up, 1 client can have several previous applications, and each of those previous applications can have several records about instalments, payments, days past due, etc.</p>",
      "rawMarkdown": "Hi Fang,\n\nlet me try to answer by example: goal of this competition is to predict risk quality of clients (or to be precise applications of clients), so client and his/her application is entity (case_id) for which your model predict probability of default (credit score).\n\n- depth=0 (what we call static attributes) are attributes that are aggregated on case_id level. Example can be age of client or gender, 1 records per 1 case_id.\n\n- depth=1, those are attributes where we have several records per client/application. Example can be previous applications or loans in credit bureau register, each client can have from 0 to n records. Therefore, for one case_id there might be several records, and to index them we use num_group1\n\n- depth=2: For some attributes with depth=1 we have more detailed information, for example for previous applications we have data about instalments like date of payments or days past due of each payments. It means for each previous application, you can have 0...n records about instalments/payments. And as an index is used num_group2\n\nTo sum up, 1 client can have several previous applications, and each of those previous applications can have several records about instalments, payments, days past due, etc.",
      "votes": null
    },
    {
      "id": "2653081",
      "postDate": "02/15/2024 07:18:00",
      "content": "<p>Got it. Thank you so much!😃</p>",
      "rawMarkdown": "Got it. Thank you so much!😃",
      "votes": null
    },
    {
      "id": "2662791",
      "postDate": "02/22/2024 06:02:09",
      "content": "<p>So that's it</p>",
      "rawMarkdown": "So that's it",
      "votes": null
    },
    {
      "id": "2667286",
      "postDate": "02/25/2024 03:44:29",
      "content": "<p>Thanks for the clarification. I have a follow up question here: for those observations where each client has record from 0 to n in \"num_group1\" (e.g., num_group1 can be 0, 1, 2, 3, etc. but the date is exactly same for all those num_group1 records), is those numeric numbers of 0 to n having some meanings such as smaller number indicates earlier record and larger number indicates later record?</p>",
      "rawMarkdown": "Thanks for the clarification. I have a follow up question here: for those observations where each client has record from 0 to n in \"num_group1\" (e.g., num_group1 can be 0, 1, 2, 3, etc. but the date is exactly same for all those num_group1 records), is those numeric numbers of 0 to n having some meanings such as smaller number indicates earlier record and larger number indicates later record?",
      "votes": null
    },
    {
      "id": "2732112",
      "postDate": "04/03/2024 02:32:38",
      "content": "<p>I have a question to ask you. </p>\n<p>You used the expression \"previous application\", what exactly does \"previous\" mean? </p>\n<p>Does that mean the application can be submitted multiple times or modified?</p>",
      "rawMarkdown": "I have a question to ask you. \n\nYou used the expression \"previous application\", what exactly does \"previous\" mean? \n\nDoes that mean the application can be submitted multiple times or modified?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2650538,
      "author_name": "tomasjeline2",
      "author_url": "",
      "post_date": "02/13/2024 14:09:18",
      "content": "<p>Hi Fang,</p>\n<p>let me try to answer by example: goal of this competition is to predict risk quality of clients (or to be precise applications of clients), so client and his/her application is entity (case_id) for which your model predict probability of default (credit score).</p>\n<ul>\n<li><p>depth=0 (what we call static attributes) are attributes that are aggregated on case_id level. Example can be age of client or gender, 1 records per 1 case_id.</p></li>\n<li><p>depth=1, those are attributes where we have several records per client/application. Example can be previous applications or loans in credit bureau register, each client can have from 0 to n records. Therefore, for one case_id there might be several records, and to index them we use num_group1</p></li>\n<li><p>depth=2: For some attributes with depth=1 we have more detailed information, for example for previous applications we have data about instalments like date of payments or days past due of each payments. It means for each previous application, you can have 0…n records about instalments/payments. And as an index is used num_group2</p></li>\n</ul>\n<p>To sum up, 1 client can have several previous applications, and each of those previous applications can have several records about instalments, payments, days past due, etc.</p>",
      "votes": null,
      "replies": [
        {
          "id": 2653081,
          "author_name": "fangshenyu",
          "author_url": "",
          "post_date": "02/15/2024 07:18:00",
          "content": "<p>Got it. Thank you so much!😃</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 2662791,
          "author_name": "ziyidong",
          "author_url": "",
          "post_date": "02/22/2024 06:02:09",
          "content": "<p>So that's it</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 2667286,
          "author_name": "haozhang16",
          "author_url": "",
          "post_date": "02/25/2024 03:44:29",
          "content": "<p>Thanks for the clarification. I have a follow up question here: for those observations where each client has record from 0 to n in \"num_group1\" (e.g., num_group1 can be 0, 1, 2, 3, etc. but the date is exactly same for all those num_group1 records), is those numeric numbers of 0 to n having some meanings such as smaller number indicates earlier record and larger number indicates later record?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 2732112,
          "author_name": "dirhq88",
          "author_url": "",
          "post_date": "04/03/2024 02:32:38",
          "content": "<p>I have a question to ask you. </p>\n<p>You used the expression \"previous application\", what exactly does \"previous\" mean? </p>\n<p>Does that mean the application can be submitted multiple times or modified?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2650392": "I can't understand what is the meaning for num_group1 and num_group2.🫠\nAnyone can help me, thanks a lot.",
    "2650538": "Hi Fang,\n\nlet me try to answer by example: goal of this competition is to predict risk quality of clients (or to be precise applications of clients), so client and his/her application is entity (case_id) for which your model predict probability of default (credit score).\n\n- depth=0 (what we call static attributes) are attributes that are aggregated on case_id level. Example can be age of client or gender, 1 records per 1 case_id.\n\n- depth=1, those are attributes where we have several records per client/application. Example can be previous applications or loans in credit bureau register, each client can have from 0 to n records. Therefore, for one case_id there might be several records, and to index them we use num_group1\n\n- depth=2: For some attributes with depth=1 we have more detailed information, for example for previous applications we have data about instalments like date of payments or days past due of each payments. It means for each previous application, you can have 0...n records about instalments/payments. And as an index is used num_group2\n\nTo sum up, 1 client can have several previous applications, and each of those previous applications can have several records about instalments, payments, days past due, etc.",
    "2653081": "Got it. Thank you so much!😃",
    "2662791": "So that's it",
    "2667286": "Thanks for the clarification. I have a follow up question here: for those observations where each client has record from 0 to n in \"num_group1\" (e.g., num_group1 can be 0, 1, 2, 3, etc. but the date is exactly same for all those num_group1 records), is those numeric numbers of 0 to n having some meanings such as smaller number indicates earlier record and larger number indicates later record?",
    "2732112": "I have a question to ask you. \n\nYou used the expression \"previous application\", what exactly does \"previous\" mean? \n\nDoes that mean the application can be submitted multiple times or modified?"
  },
  "source": "meta"
}