{
  "id": 329467,
  "title": "Feature 'S_2' (Monthly Statement date): initial analysis and engineering",
  "url": "/competitions/amex-default-prediction/discussion/329467",
  "author_name": "",
  "post_date": "2022-06-06T20:25:11.331005900Z",
  "votes": 15,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I created a quick notebook to look at S_2 under the hood, as well as show one way to convert it to numeric for feature engineering.</p>\n<p><a href=\"https://www.kaggle.com/code/roberthatch/amex-s2-feature-engg\" target=\"_blank\">https://www.kaggle.com/code/roberthatch/amex-s2-feature-engg</a></p>\n<p>Credit: Notebook was built on top of <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> <a href=\"https://www.kaggle.com/raddar\" target=\"_blank\">@raddar</a> <a href=\"https://www.kaggle.com/huseyincot\" target=\"_blank\">@huseyincot</a> 's work.</p>\n<p>Of note: as expected, short-term customers (&lt;13 monthly statements, all statements consecutive) are much higher risk (43%), but gap customers (&lt;13 monthly statements, not all consecutive) aren't particularly (27%). With long-term customers having (23%) risk. [Numbers are ignoring the 20x multiplier of course.]</p>\n<p>I also thought it was strange that short-term customers are MORE risk the longer they've been a customer, which seems very backwards, but maybe there's a seasonal component, and from April (taxes?) to September might be more prone to higher risk activities? Anyways, that was interesting.</p>\n<p>Anyone else looked into this column? What have you discovered?</p>",
  "messages": [
    {
      "id": "1813444",
      "postDate": "06/06/2022 20:25:11",
      "content": "<p>I created a quick notebook to look at S_2 under the hood, as well as show one way to convert it to numeric for feature engineering.</p>\n<p><a href=\"https://www.kaggle.com/code/roberthatch/amex-s2-feature-engg\" target=\"_blank\">https://www.kaggle.com/code/roberthatch/amex-s2-feature-engg</a></p>\n<p>Credit: Notebook was built on top of <a href=\"https://www.kaggle.com/cdeotte\" target=\"_blank\">@cdeotte</a> <a href=\"https://www.kaggle.com/raddar\" target=\"_blank\">@raddar</a> <a href=\"https://www.kaggle.com/huseyincot\" target=\"_blank\">@huseyincot</a> 's work.</p>\n<p>Of note: as expected, short-term customers (&lt;13 monthly statements, all statements consecutive) are much higher risk (43%), but gap customers (&lt;13 monthly statements, not all consecutive) aren't particularly (27%). With long-term customers having (23%) risk. [Numbers are ignoring the 20x multiplier of course.]</p>\n<p>I also thought it was strange that short-term customers are MORE risk the longer they've been a customer, which seems very backwards, but maybe there's a seasonal component, and from April (taxes?) to September might be more prone to higher risk activities? Anyways, that was interesting.</p>\n<p>Anyone else looked into this column? What have you discovered?</p>",
      "rawMarkdown": "I created a quick notebook to look at S_2 under the hood, as well as show one way to convert it to numeric for feature engineering.\n\nhttps://www.kaggle.com/code/roberthatch/amex-s2-feature-engg\n\nCredit: Notebook was built on top of @cdeotte @raddar @huseyincot 's work.\n\nOf note: as expected, short-term customers (<13 monthly statements, all statements consecutive) are much higher risk (43%), but gap customers (<13 monthly statements, not all consecutive) aren't particularly (27%). With long-term customers having (23%) risk. [Numbers are ignoring the 20x multiplier of course.]\n\nI also thought it was strange that short-term customers are MORE risk the longer they've been a customer, which seems very backwards, but maybe there's a seasonal component, and from April (taxes?) to September might be more prone to higher risk activities? Anyways, that was interesting.\n\nAnyone else looked into this column? What have you discovered?",
      "votes": null
    },
    {
      "id": "1895417",
      "postDate": "08/12/2022 05:58:20",
      "content": "<p>Interesting observation!</p>",
      "rawMarkdown": "Interesting observation!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1895417,
      "author_name": "kimberlynie",
      "author_url": "",
      "post_date": "08/12/2022 05:58:20",
      "content": "<p>Interesting observation!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1813444": "I created a quick notebook to look at S_2 under the hood, as well as show one way to convert it to numeric for feature engineering.\n\nhttps://www.kaggle.com/code/roberthatch/amex-s2-feature-engg\n\nCredit: Notebook was built on top of @cdeotte @raddar @huseyincot 's work.\n\nOf note: as expected, short-term customers (<13 monthly statements, all statements consecutive) are much higher risk (43%), but gap customers (<13 monthly statements, not all consecutive) aren't particularly (27%). With long-term customers having (23%) risk. [Numbers are ignoring the 20x multiplier of course.]\n\nI also thought it was strange that short-term customers are MORE risk the longer they've been a customer, which seems very backwards, but maybe there's a seasonal component, and from April (taxes?) to September might be more prone to higher risk activities? Anyways, that was interesting.\n\nAnyone else looked into this column? What have you discovered?",
    "1895417": "Interesting observation!"
  },
  "source": "meta"
}