{
  "id": 503071,
  "title": "Trying to understand P transformation",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/503071",
  "author_name": "Andrey Zhuravlev",
  "post_date": "2024-05-15T23:42:19.715000",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hello, I'm struggling to understand DPD (Days Past Due) ending with a 'P' suffix variables that have been transformed.</p>\n<p>For example, in credit_bureau_b_1 table, dpd550P (The number of days past due for active loans where a guarantee has been provided) has a bunch of values in 100s of thousands and even millions, pmtdaysoverdue_1135P (Number of days past due for existing contracts in the credit bureau) in the same table has values in 100s and 1000s, but there are 4 observations that are over 600k. dpdmax_851P also has values in 10s of thousands, as well as millions.  They are also precise to the last decimal point (for example 1014455.0 in dpdmax_851P).</p>\n<p>I am far from an expert in data science, can you please help me understand what kind of transformations these could be and what is the possible reason for them to be used? Are such transformations typical for day-based metrics?</p>",
  "messages": [
    {
      "id": 2815532,
      "postDate": "2024-05-15T23:42:19.717Z",
      "content": "<p>Hello, I'm struggling to understand DPD (Days Past Due) ending with a 'P' suffix variables that have been transformed.</p>\n<p>For example, in credit_bureau_b_1 table, dpd550P (The number of days past due for active loans where a guarantee has been provided) has a bunch of values in 100s of thousands and even millions, pmtdaysoverdue_1135P (Number of days past due for existing contracts in the credit bureau) in the same table has values in 100s and 1000s, but there are 4 observations that are over 600k. dpdmax_851P also has values in 10s of thousands, as well as millions.  They are also precise to the last decimal point (for example 1014455.0 in dpdmax_851P).</p>\n<p>I am far from an expert in data science, can you please help me understand what kind of transformations these could be and what is the possible reason for them to be used? Are such transformations typical for day-based metrics?</p>",
      "rawMarkdown": "Hello, I'm struggling to understand DPD (Days Past Due) ending with a 'P' suffix variables that have been transformed.\n\nFor example, in credit_bureau_b_1 table, dpd550P (The number of days past due for active loans where a guarantee has been provided) has a bunch of values in 100s of thousands and even millions, pmtdaysoverdue_1135P (Number of days past due for existing contracts in the credit bureau) in the same table has values in 100s and 1000s, but there are 4 observations that are over 600k. dpdmax_851P also has values in 10s of thousands, as well as millions.  They are also precise to the last decimal point (for example 1014455.0 in dpdmax_851P).\n\nI am far from an expert in data science, can you please help me understand what kind of transformations these could be and what is the possible reason for them to be used? Are such transformations typical for day-based metrics?",
      "votes": 1
    },
    {
      "id": 2818071,
      "postDate": "2024-05-17T09:03:42.730Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2818071,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-05-17T09:03:42.730000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2815532": "Hello, I'm struggling to understand DPD (Days Past Due) ending with a 'P' suffix variables that have been transformed.\n\nFor example, in credit_bureau_b_1 table, dpd550P (The number of days past due for active loans where a guarantee has been provided) has a bunch of values in 100s of thousands and even millions, pmtdaysoverdue_1135P (Number of days past due for existing contracts in the credit bureau) in the same table has values in 100s and 1000s, but there are 4 observations that are over 600k. dpdmax_851P also has values in 10s of thousands, as well as millions.  They are also precise to the last decimal point (for example 1014455.0 in dpdmax_851P).\n\nI am far from an expert in data science, can you please help me understand what kind of transformations these could be and what is the possible reason for them to be used? Are such transformations typical for day-based metrics?",
    "2818071": ""
  }
}