{
  "id": 39913,
  "title": "Please explain \"user does not falls in our scope of prediction\"",
  "url": "/competitions/kkbox-churn-prediction-challenge/discussion/39913",
  "author_name": "",
  "post_date": "2017-09-23T20:19:10.801453300Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Why we are not predicting the is_churn of a user if its expiration date does not fall in between 2017-03-01 to 2017-03-31? Please explain this.</p>",
  "messages": [
    {
      "id": "223842",
      "postDate": "09/23/2017 20:19:10",
      "content": "<p>Why we are not predicting the is_churn of a user if its expiration date does not fall in between 2017-03-01 to 2017-03-31? Please explain this.</p>",
      "rawMarkdown": "Why we are not predicting the is_churn of a user if its expiration date does not fall in between 2017-03-01 to 2017-03-31? Please explain this.",
      "votes": null
    },
    {
      "id": "224078",
      "postDate": "09/25/2017 03:21:58",
      "content": "<p>Those two dates is just a criteria for selecting users.</p>",
      "rawMarkdown": "Those two dates is just a criteria for selecting users.",
      "votes": null
    },
    {
      "id": "224199",
      "postDate": "09/25/2017 14:05:10",
      "content": "<p>If one user has transaction logs like (20170104,20170204, false ), (20170115,20170304, false ),\nbecause in the first entry, the user's membership_expire_date = 20170204  falls  in Feb., \nis it a training sample? Is it a training sample, what is value of is_churn, is a churn user or not?</p>\n\n<p>Thank you!\n@Arden Chiu</p>",
      "rawMarkdown": "If one user has transaction logs like (20170104,20170204, false ), (20170115,20170304, false ),\nbecause in the first entry, the user's membership_expire_date = 20170204  falls  in Feb., \nis it a training sample? Is it a training sample, what is value of is_churn, is a churn user or not?\n\nThank you!\n@Arden Chiu",
      "votes": null
    },
    {
      "id": "224396",
      "postDate": "09/26/2017 06:06:28",
      "content": "<p>For training, you take a snapshot of the transaction logs . Depending on the date of the snapshot, you will have two cases:\n1. (20170115,20170304, false) in your snapshot: you don't need to predict this user as he extend his membership prior to expiration date. e.g. your snapshot taken between (20170105 - 20170114)\n2.(20170115,20170304, false) is not in your snapshot: you take a peek into the future and found the entry.\nis_churn = ~~false~~ true because you will not find any subscription log within 30 days after 20170204. </p>",
      "rawMarkdown": "For training, you take a snapshot of the transaction logs . Depending on the date of the snapshot, you will have two cases:\n1. (20170115,20170304, false) in your snapshot: you don't need to predict this user as he extend his membership prior to expiration date. e.g. your snapshot taken between (20170105 - 20170114)\n2.(20170115,20170304, false) is not in your snapshot: you take a peek into the future and found the entry.\nis_churn = ~~false~~ true because you will not find any subscription log within 30 days after 20170204.",
      "votes": null
    },
    {
      "id": "224552",
      "postDate": "09/26/2017 18:41:57",
      "content": "<p>It would be is_chur=true, isn't it? Because \"not find any subscription log within 30 days after expiration date\" is the definition of is_churn=true.</p>",
      "rawMarkdown": "It would be is_chur=true, isn't it? Because \"not find any subscription log within 30 days after expiration date\" is the definition of is_churn=true.",
      "votes": null
    },
    {
      "id": "224612",
      "postDate": "09/27/2017 02:22:36",
      "content": "<p>Correct, Thanks for pointing out my typo, corrected the error in original post</p>",
      "rawMarkdown": "Correct, Thanks for pointing out my typo, corrected the error in original post",
      "votes": null
    },
    {
      "id": "224793",
      "postDate": "09/27/2017 17:11:34",
      "content": "<p>I don't understand well the idea of the snapshot. We must discover when is it taken in the train and tests dataset in order to build a more extensive train dataset? </p>\n\n<p>Thank you! @Arden Chiu</p>",
      "rawMarkdown": "I don't understand well the idea of the snapshot. We must discover when is it taken in the train and tests dataset in order to build a more extensive train dataset? \n\nThank you! @Arden Chiu",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 224078,
      "author_name": "ardenkkbox",
      "author_url": "",
      "post_date": "09/25/2017 03:21:58",
      "content": "<p>Those two dates is just a criteria for selecting users.</p>",
      "votes": null,
      "replies": [
        {
          "id": 224199,
          "author_name": "huoyangyang",
          "author_url": "",
          "post_date": "09/25/2017 14:05:10",
          "content": "<p>If one user has transaction logs like (20170104,20170204, false ), (20170115,20170304, false ),\nbecause in the first entry, the user's membership_expire_date = 20170204  falls  in Feb., \nis it a training sample? Is it a training sample, what is value of is_churn, is a churn user or not?</p>\n\n<p>Thank you!\n@Arden Chiu</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 224396,
          "author_name": "ardenkkbox",
          "author_url": "",
          "post_date": "09/26/2017 06:06:28",
          "content": "<p>For training, you take a snapshot of the transaction logs . Depending on the date of the snapshot, you will have two cases:\n1. (20170115,20170304, false) in your snapshot: you don't need to predict this user as he extend his membership prior to expiration date. e.g. your snapshot taken between (20170105 - 20170114)\n2.(20170115,20170304, false) is not in your snapshot: you take a peek into the future and found the entry.\nis_churn = ~~false~~ true because you will not find any subscription log within 30 days after 20170204. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 224552,
          "author_name": "joulazas",
          "author_url": "",
          "post_date": "09/26/2017 18:41:57",
          "content": "<p>It would be is_chur=true, isn't it? Because \"not find any subscription log within 30 days after expiration date\" is the definition of is_churn=true.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 224612,
          "author_name": "ardenkkbox",
          "author_url": "",
          "post_date": "09/27/2017 02:22:36",
          "content": "<p>Correct, Thanks for pointing out my typo, corrected the error in original post</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 224793,
          "author_name": "joulazas",
          "author_url": "",
          "post_date": "09/27/2017 17:11:34",
          "content": "<p>I don't understand well the idea of the snapshot. We must discover when is it taken in the train and tests dataset in order to build a more extensive train dataset? </p>\n\n<p>Thank you! @Arden Chiu</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "223842": "Why we are not predicting the is_churn of a user if its expiration date does not fall in between 2017-03-01 to 2017-03-31? Please explain this.",
    "224078": "Those two dates is just a criteria for selecting users.",
    "224199": "If one user has transaction logs like (20170104,20170204, false ), (20170115,20170304, false ),\nbecause in the first entry, the user's membership_expire_date = 20170204  falls  in Feb., \nis it a training sample? Is it a training sample, what is value of is_churn, is a churn user or not?\n\nThank you!\n@Arden Chiu",
    "224396": "For training, you take a snapshot of the transaction logs . Depending on the date of the snapshot, you will have two cases:\n1. (20170115,20170304, false) in your snapshot: you don't need to predict this user as he extend his membership prior to expiration date. e.g. your snapshot taken between (20170105 - 20170114)\n2.(20170115,20170304, false) is not in your snapshot: you take a peek into the future and found the entry.\nis_churn = ~~false~~ true because you will not find any subscription log within 30 days after 20170204.",
    "224552": "It would be is_chur=true, isn't it? Because \"not find any subscription log within 30 days after expiration date\" is the definition of is_churn=true.",
    "224612": "Correct, Thanks for pointing out my typo, corrected the error in original post",
    "224793": "I don't understand well the idea of the snapshot. We must discover when is it taken in the train and tests dataset in order to build a more extensive train dataset? \n\nThank you! @Arden Chiu"
  },
  "source": "meta"
}