{
  "id": 331634,
  "title": "Cohorts explained",
  "url": "/competitions/amex-default-prediction/discussion/331634",
  "author_name": "",
  "post_date": "2022-06-18T09:03:48.957372600Z",
  "votes": 14,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I created a notebooks which revealed some time patterns in almost all features.</p>\n<p>I just want to explain some related key points:<br>\nWe got part of the data which we (in b2b2c industries like marketing) call cohort - customers which have started their customer journey together (Mar-2017). When they will be about 6 month old in Sep-2017 - their features will be totally different from the features of 3-month-old customers which started in June-2017. This is why you can see spikes and change in behavior between train and test.<br>\nWe have only 1 cohort in train and 2 cohorts in test.</p>\n<p>More advanced explanation can be found on internet, for example <a href=\"https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing\" target=\"_blank\">https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing</a></p>",
  "messages": [
    {
      "id": "1824381",
      "postDate": "06/18/2022 09:03:48",
      "content": "<p>I created a notebooks which revealed some time patterns in almost all features.</p>\n<p>I just want to explain some related key points:<br>\nWe got part of the data which we (in b2b2c industries like marketing) call cohort - customers which have started their customer journey together (Mar-2017). When they will be about 6 month old in Sep-2017 - their features will be totally different from the features of 3-month-old customers which started in June-2017. This is why you can see spikes and change in behavior between train and test.<br>\nWe have only 1 cohort in train and 2 cohorts in test.</p>\n<p>More advanced explanation can be found on internet, for example <a href=\"https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing\" target=\"_blank\">https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing</a></p>",
      "rawMarkdown": "I created a notebooks which revealed some time patterns in almost all features.\n\nI just want to explain some related key points:\nWe got part of the data which we (in b2b2c industries like marketing) call cohort - customers which have started their customer journey together (Mar-2017). When they will be about 6 month old in Sep-2017 - their features will be totally different from the features of 3-month-old customers which started in June-2017. This is why you can see spikes and change in behavior between train and test.\nWe have only 1 cohort in train and 2 cohorts in test.\n\nMore advanced explanation can be found on internet, for example https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing",
      "votes": null
    },
    {
      "id": "1838950",
      "postDate": "07/01/2022 02:49:58",
      "content": "<p>Thank you for sharing your work. </p>",
      "rawMarkdown": "Thank you for sharing your work.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1838950,
      "author_name": "mohammadrahmati",
      "author_url": "",
      "post_date": "07/01/2022 02:49:58",
      "content": "<p>Thank you for sharing your work. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1824381": "I created a notebooks which revealed some time patterns in almost all features.\n\nI just want to explain some related key points:\nWe got part of the data which we (in b2b2c industries like marketing) call cohort - customers which have started their customer journey together (Mar-2017). When they will be about 6 month old in Sep-2017 - their features will be totally different from the features of 3-month-old customers which started in June-2017. This is why you can see spikes and change in behavior between train and test.\nWe have only 1 cohort in train and 2 cohorts in test.\n\nMore advanced explanation can be found on internet, for example https://www.adroll.com/blog/a-beginner-guide-to-cohort-audiences-in-marketing",
    "1838950": "Thank you for sharing your work."
  },
  "source": "meta"
}