{
  "id": 485634,
  "title": "Clarification on train_credit_bureau_b_2 dataset regarding num_group1 and DPD calculations",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/485634",
  "author_name": "",
  "post_date": "2024-03-21T14:52:52.330786800Z",
  "votes": 3,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hello everyone,</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3137068%2F642cfb667e886acae12afaae0cde6efb%2F2024-03-21%2023.51.17.png?generation=1711032696799569&amp;alt=media\" alt=\"train_credit_bureau_b_2 data\"></p>\n<p>I am currently working with the train_credit_bureau_b_2 dataset for case_id=1702371（I have attached a portion of the data.） and have a couple of questions regarding the interpretation of num_group1, num_group2, and the DPD values. I would really appreciate any insights or explanations from the community.</p>\n<p>In the dataset, there are instances where num_group1 is either 0 or 1. Assuming num_group1=0 represents contracts by the applicant themselves, what does num_group1=1 signify? Is it related to co-signed or guarantor-backed contracts, or does it indicate something else?</p>\n<p>For a given contract (identified by num_group1), there's a particular record where num_group2=2 has a DPD (Days Past Due) value of 9186 days. Since num_group2 delineates individual payment instances for the same contract, I would expect the DPD values for num_group2=0 and num_group2=1 to logically precede this, potentially showing 9186 minus one or two months, respectively. Is the DPD value only recorded at the point of payment occurrence? Does a value of 0 imply no payment was due at that time?</p>\n<p>Regarding the pmts_pmtsoverdue_635A column, which indicates the existence and extent of payment overdue, how is the \"extent\" quantified? In this dataset, we see a value of 0.2 for this field. What scale or measure is used to determine this number?</p>\n<p>Thank you in advance for your help!</p>",
  "messages": [
    {
      "id": "2709306",
      "postDate": "03/21/2024 14:52:52",
      "content": "<p>Hello everyone,</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3137068%2F642cfb667e886acae12afaae0cde6efb%2F2024-03-21%2023.51.17.png?generation=1711032696799569&amp;alt=media\" alt=\"train_credit_bureau_b_2 data\"></p>\n<p>I am currently working with the train_credit_bureau_b_2 dataset for case_id=1702371（I have attached a portion of the data.） and have a couple of questions regarding the interpretation of num_group1, num_group2, and the DPD values. I would really appreciate any insights or explanations from the community.</p>\n<p>In the dataset, there are instances where num_group1 is either 0 or 1. Assuming num_group1=0 represents contracts by the applicant themselves, what does num_group1=1 signify? Is it related to co-signed or guarantor-backed contracts, or does it indicate something else?</p>\n<p>For a given contract (identified by num_group1), there's a particular record where num_group2=2 has a DPD (Days Past Due) value of 9186 days. Since num_group2 delineates individual payment instances for the same contract, I would expect the DPD values for num_group2=0 and num_group2=1 to logically precede this, potentially showing 9186 minus one or two months, respectively. Is the DPD value only recorded at the point of payment occurrence? Does a value of 0 imply no payment was due at that time?</p>\n<p>Regarding the pmts_pmtsoverdue_635A column, which indicates the existence and extent of payment overdue, how is the \"extent\" quantified? In this dataset, we see a value of 0.2 for this field. What scale or measure is used to determine this number?</p>\n<p>Thank you in advance for your help!</p>",
      "rawMarkdown": "Hello everyone,\n\n![train_credit_bureau_b_2 data](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3137068%2F642cfb667e886acae12afaae0cde6efb%2F2024-03-21%2023.51.17.png?generation=1711032696799569&alt=media)\n\nI am currently working with the train_credit_bureau_b_2 dataset for case_id=1702371（I have attached a portion of the data.） and have a couple of questions regarding the interpretation of num_group1, num_group2, and the DPD values. I would really appreciate any insights or explanations from the community.\n\nIn the dataset, there are instances where num_group1 is either 0 or 1. Assuming num_group1=0 represents contracts by the applicant themselves, what does num_group1=1 signify? Is it related to co-signed or guarantor-backed contracts, or does it indicate something else?\n\nFor a given contract (identified by num_group1), there's a particular record where num_group2=2 has a DPD (Days Past Due) value of 9186 days. Since num_group2 delineates individual payment instances for the same contract, I would expect the DPD values for num_group2=0 and num_group2=1 to logically precede this, potentially showing 9186 minus one or two months, respectively. Is the DPD value only recorded at the point of payment occurrence? Does a value of 0 imply no payment was due at that time?\n\nRegarding the pmts_pmtsoverdue_635A column, which indicates the existence and extent of payment overdue, how is the \"extent\" quantified? In this dataset, we see a value of 0.2 for this field. What scale or measure is used to determine this number?\n\nThank you in advance for your help!",
      "votes": null
    },
    {
      "id": "2709809",
      "postDate": "03/21/2024 21:02:42",
      "content": "<ul>\n<li>the num_group1 in  credit_bureau_b_2 referes to the num_group1  in credit_bureau_b_1.</li>\n<li>Its all client credits. In your case 1702371 the client has only two documented credits</li>\n<li>Look at e.g. case 1445 to see a different scenario with 0-4 in num_group1</li>\n<li>the DPD value of 9186 is rather the monetary value which was paid late</li>\n<li>the 0.2 are rather the percentage of the total credit value. the total credit value is documented in credit_bureau_b_1 column amount_1115A with 43998.0. 9186 is about 20% of that value. I rather trust the data then the documentation here.</li>\n</ul>",
      "rawMarkdown": "the num_group1 in  credit_bureau_b_2 referes to the num_group1  in credit_bureau_b_1.\n- Its all client credits. In your case 1702371 the client has only two documented credits\n- Look at e.g. case 1445 to see a different scenario with 0-4 in num_group1\n- the DPD value of 9186 is rather the monetary value which was paid late\n- the 0.2 are rather the percentage of the total credit value. the total credit value is documented in credit_bureau_b_1 column amount_1115A with 43998.0. 9186 is about 20% of that value. I rather trust the data then the documentation here.",
      "votes": null
    },
    {
      "id": "2710030",
      "postDate": "03/22/2024 00:30:37",
      "content": "<p>Thank you! I completely understand!</p>",
      "rawMarkdown": "Thank you! I completely understand!",
      "votes": null
    },
    {
      "id": "2761194",
      "postDate": "04/19/2024 17:36:24",
      "content": "<p>In case 1445, I see some num_group2 with 24,26 th value which is kinda opposite, they are almost same but pmts_overdue_635A is decreased even when the pmts_dpdvalue_108P increased</p>\n<p>Can you elaborate on this? I think I am missing something, thanks in advance</p>",
      "rawMarkdown": "In case 1445, I see some num_group2 with 24,26 th value which is kinda opposite, they are almost same but pmts_overdue_635A is decreased even when the pmts_dpdvalue_108P increased\n\nCan you elaborate on this? I think I am missing something, thanks in advance",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2709809,
      "author_name": "aahhammer",
      "author_url": "",
      "post_date": "03/21/2024 21:02:42",
      "content": "<ul>\n<li>the num_group1 in  credit_bureau_b_2 referes to the num_group1  in credit_bureau_b_1.</li>\n<li>Its all client credits. In your case 1702371 the client has only two documented credits</li>\n<li>Look at e.g. case 1445 to see a different scenario with 0-4 in num_group1</li>\n<li>the DPD value of 9186 is rather the monetary value which was paid late</li>\n<li>the 0.2 are rather the percentage of the total credit value. the total credit value is documented in credit_bureau_b_1 column amount_1115A with 43998.0. 9186 is about 20% of that value. I rather trust the data then the documentation here.</li>\n</ul>",
      "votes": null,
      "replies": [
        {
          "id": 2710030,
          "author_name": "pandaman817",
          "author_url": "",
          "post_date": "03/22/2024 00:30:37",
          "content": "<p>Thank you! I completely understand!</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 2761194,
          "author_name": "harshavardhane",
          "author_url": "",
          "post_date": "04/19/2024 17:36:24",
          "content": "<p>In case 1445, I see some num_group2 with 24,26 th value which is kinda opposite, they are almost same but pmts_overdue_635A is decreased even when the pmts_dpdvalue_108P increased</p>\n<p>Can you elaborate on this? I think I am missing something, thanks in advance</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "2709306": "Hello everyone,\n\n![train_credit_bureau_b_2 data](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3137068%2F642cfb667e886acae12afaae0cde6efb%2F2024-03-21%2023.51.17.png?generation=1711032696799569&alt=media)\n\nI am currently working with the train_credit_bureau_b_2 dataset for case_id=1702371（I have attached a portion of the data.） and have a couple of questions regarding the interpretation of num_group1, num_group2, and the DPD values. I would really appreciate any insights or explanations from the community.\n\nIn the dataset, there are instances where num_group1 is either 0 or 1. Assuming num_group1=0 represents contracts by the applicant themselves, what does num_group1=1 signify? Is it related to co-signed or guarantor-backed contracts, or does it indicate something else?\n\nFor a given contract (identified by num_group1), there's a particular record where num_group2=2 has a DPD (Days Past Due) value of 9186 days. Since num_group2 delineates individual payment instances for the same contract, I would expect the DPD values for num_group2=0 and num_group2=1 to logically precede this, potentially showing 9186 minus one or two months, respectively. Is the DPD value only recorded at the point of payment occurrence? Does a value of 0 imply no payment was due at that time?\n\nRegarding the pmts_pmtsoverdue_635A column, which indicates the existence and extent of payment overdue, how is the \"extent\" quantified? In this dataset, we see a value of 0.2 for this field. What scale or measure is used to determine this number?\n\nThank you in advance for your help!",
    "2709809": "the num_group1 in  credit_bureau_b_2 referes to the num_group1  in credit_bureau_b_1.\n- Its all client credits. In your case 1702371 the client has only two documented credits\n- Look at e.g. case 1445 to see a different scenario with 0-4 in num_group1\n- the DPD value of 9186 is rather the monetary value which was paid late\n- the 0.2 are rather the percentage of the total credit value. the total credit value is documented in credit_bureau_b_1 column amount_1115A with 43998.0. 9186 is about 20% of that value. I rather trust the data then the documentation here.",
    "2710030": "Thank you! I completely understand!",
    "2761194": "In case 1445, I see some num_group2 with 24,26 th value which is kinda opposite, they are almost same but pmts_overdue_635A is decreased even when the pmts_dpdvalue_108P increased\n\nCan you elaborate on this? I think I am missing something, thanks in advance"
  },
  "source": "meta"
}