{
  "id": 327595,
  "title": "Mutiple vintages for test data - April 2019 and Oct 2019",
  "url": "/competitions/amex-default-prediction/discussion/327595",
  "author_name": "Seeker",
  "post_date": "2022-05-28T03:34:26.855000",
  "votes": 6,
  "comment_count": 5,
  "views": 0,
  "content": "<p><strong>Train data is for single (March 2018) Vintage and test data is for multiple vintages</strong>.<br>\n For details check this <a href=\"https://www.kaggle.com/code/taran8727/exploration-time-periods-and-credit-defaults\" target=\"_blank\">notebook</a></p>\n<p>Since credit defaults are affected by macro economic conditions (inflation,stimulus), Single vintage in train set will make generalization challenging. </p>\n<p><a href=\"https://fred.stlouisfed.org/series/DRCCLACBS\" target=\"_blank\">FRED data Defaults over time</a></p>\n<p>Default rates are generally lower for Feb to April, due to factors such as Tax refunds, so error might be higher on October vintage<strong><em><em></em></em></strong></p>",
  "messages": [
    {
      "id": 1803642,
      "postDate": "2022-05-28T03:34:26.857Z",
      "content": "<p><strong>Train data is for single (March 2018) Vintage and test data is for multiple vintages</strong>.<br>\n For details check this <a href=\"https://www.kaggle.com/code/taran8727/exploration-time-periods-and-credit-defaults\" target=\"_blank\">notebook</a></p>\n<p>Since credit defaults are affected by macro economic conditions (inflation,stimulus), Single vintage in train set will make generalization challenging. </p>\n<p><a href=\"https://fred.stlouisfed.org/series/DRCCLACBS\" target=\"_blank\">FRED data Defaults over time</a></p>\n<p>Default rates are generally lower for Feb to April, due to factors such as Tax refunds, so error might be higher on October vintage<strong><em><em></em></em></strong></p>",
      "rawMarkdown": "**Train data is for single (March 2018) Vintage and test data is for multiple vintages**.\n For details check this [notebook](https://www.kaggle.com/code/taran8727/exploration-time-periods-and-credit-defaults)\n\nSince credit defaults are affected by macro economic conditions (inflation,stimulus), Single vintage in train set will make generalization challenging. \n\n[FRED data Defaults over time](https://fred.stlouisfed.org/series/DRCCLACBS)\n\nDefault rates are generally lower for Feb to April, due to factors such as Tax refunds, so error might be higher on October vintage********\n\n",
      "votes": 5
    },
    {
      "id": 1806730,
      "postDate": "2022-05-31T12:32:34.497Z",
      "content": "<p>I'm still trying to understand the problem. If I understood correctly, all of the customers' last credit card statement is between 2018-03-01 and 2018-03-31 in training set. If they don't pay back their credit card balance amount in 120 days after the last credit card statement, then target value becomes 1. For test set, customers' last credit card statement can be either between in 2018-04-01 and 2018-04-30 or 2018-10-01 and 2018-10-31.</p>",
      "rawMarkdown": "I'm still trying to understand the problem. If I understood correctly, all of the customers' last credit card statement is between 2018-03-01 and 2018-03-31 in training set. If they don't pay back their credit card balance amount in 120 days after the last credit card statement, then target value becomes 1. For test set, customers' last credit card statement can be either between in 2018-04-01 and 2018-04-30 or 2018-10-01 and 2018-10-31.",
      "replies": [
        {
          "id": 1807383,
          "postDate": "2022-06-01T00:46:44.230Z",
          "content": "<p>Actually its 120+ days default in next 18 months from the last statement, we are given a years (13 statements) data to predict behavior in next 18 months</p>",
          "rawMarkdown": "Actually its 120+ days default in next 18 months from the last statement, we are given a years (13 statements) data to predict behavior in next 18 months"
        },
        {
          "id": 1814522,
          "postDate": "2022-06-08T02:27:23.160Z",
          "rawMarkdown": "",
          "votes": 1,
          "isDeleted": true
        },
        {
          "id": 1820806,
          "postDate": "2022-06-15T02:29:44.573Z",
          "content": "<p>13th statement can be the first statement on which someone became delinquent in the 18 month window.<br>\nThis 18 month window starts after 13th statement.</p>\n<p>Basically from 13 months usage behavior we need to predict customers probability of 120+ default in next 18 months</p>",
          "rawMarkdown": "13th statement can be the first statement on which someone became delinquent in the 18 month window.\nThis 18 month window starts after 13th statement.\n\nBasically from 13 months usage behavior we need to predict customers probability of 120+ default in next 18 months\n"
        },
        {
          "id": 1821056,
          "postDate": "2022-06-15T08:24:03.433Z",
          "rawMarkdown": "",
          "votes": 1,
          "isDeleted": true
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 1806730,
      "author_name": "Gunes Evitan",
      "author_url": "",
      "post_date": "2022-05-31T12:32:34.497000",
      "content": "<p>I'm still trying to understand the problem. If I understood correctly, all of the customers' last credit card statement is between 2018-03-01 and 2018-03-31 in training set. If they don't pay back their credit card balance amount in 120 days after the last credit card statement, then target value becomes 1. For test set, customers' last credit card statement can be either between in 2018-04-01 and 2018-04-30 or 2018-10-01 and 2018-10-31.</p>",
      "votes": 0,
      "replies": [
        {
          "id": 1807383,
          "author_name": "Seeker",
          "author_url": "",
          "post_date": "2022-06-01T00:46:44.230000",
          "content": "<p>Actually its 120+ days default in next 18 months from the last statement, we are given a years (13 statements) data to predict behavior in next 18 months</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1814522,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-06-08T02:27:23.160000",
          "content": "",
          "votes": 1,
          "replies": []
        },
        {
          "id": 1820806,
          "author_name": "Seeker",
          "author_url": "",
          "post_date": "2022-06-15T02:29:44.573000",
          "content": "<p>13th statement can be the first statement on which someone became delinquent in the 18 month window.<br>\nThis 18 month window starts after 13th statement.</p>\n<p>Basically from 13 months usage behavior we need to predict customers probability of 120+ default in next 18 months</p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 1821056,
          "author_name": "",
          "author_url": "",
          "post_date": "2022-06-15T08:24:03.433000",
          "content": "",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1803642": "**Train data is for single (March 2018) Vintage and test data is for multiple vintages**.\n For details check this [notebook](https://www.kaggle.com/code/taran8727/exploration-time-periods-and-credit-defaults)\n\nSince credit defaults are affected by macro economic conditions (inflation,stimulus), Single vintage in train set will make generalization challenging. \n\n[FRED data Defaults over time](https://fred.stlouisfed.org/series/DRCCLACBS)\n\nDefault rates are generally lower for Feb to April, due to factors such as Tax refunds, so error might be higher on October vintage********\n\n",
    "1806730": "I'm still trying to understand the problem. If I understood correctly, all of the customers' last credit card statement is between 2018-03-01 and 2018-03-31 in training set. If they don't pay back their credit card balance amount in 120 days after the last credit card statement, then target value becomes 1. For test set, customers' last credit card statement can be either between in 2018-04-01 and 2018-04-30 or 2018-10-01 and 2018-10-31."
  }
}