{
  "id": 51190,
  "title": "Why there is a gap between training set and test set in click_time?",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/51190",
  "author_name": "",
  "post_date": "2018-03-06T11:28:03.984184800Z",
  "votes": 2,
  "comment_count": 9,
  "views": 0,
  "content": "<p>Hello, the max timestamp of training set is 2017-11-09 16:00:00 , But the min timestamp of test set is 2017-11-10 04:00:00.\nI donot know why there is a gap between training set and test set. It is really bad to have this kind of gap.\nAnyone knows? Thank you in advance.</p>",
  "messages": [
    {
      "id": "291557",
      "postDate": "03/06/2018 11:28:03",
      "content": "<p>Hello, the max timestamp of training set is 2017-11-09 16:00:00 , But the min timestamp of test set is 2017-11-10 04:00:00.\nI donot know why there is a gap between training set and test set. It is really bad to have this kind of gap.\nAnyone knows? Thank you in advance.</p>",
      "rawMarkdown": "Hello, the max timestamp of training set is 2017-11-09 16:00:00 , But the min timestamp of test set is 2017-11-10 04:00:00.\nI donot know why there is a gap between training set and test set. It is really bad to have this kind of gap.\nAnyone knows? Thank you in advance.",
      "votes": null
    },
    {
      "id": "291640",
      "postDate": "03/06/2018 14:44:51",
      "content": "<p>I believe the timestamp is in UTC, not China local timezone. \nChina local timezone is 8 hours earlier than UTC, so it's actually 2017-11-09-23:59:59 in China local timezone. \nBut the min time in test set is a little bit wired, I will have a closer look.</p>",
      "rawMarkdown": "I believe the timestamp is in UTC, not China local timezone. \nChina local timezone is 8 hours earlier than UTC, so it's actually 2017-11-09-23:59:59 in China local timezone. \nBut the min time in test set is a little bit wired, I will have a closer look.",
      "votes": null
    },
    {
      "id": "291847",
      "postDate": "03/07/2018 01:34:02",
      "content": "<p>\"But the min time in test set is a little bit wired\", That is right. </p>",
      "rawMarkdown": "\"But the min time in test set is a little bit wired\", That is right.",
      "votes": null
    },
    {
      "id": "291873",
      "postDate": "03/07/2018 02:38:16",
      "content": "<p>Is it true? if this true this data set might not be properly shuffled. </p>",
      "rawMarkdown": "Is it true? if this true this data set might not be properly shuffled.",
      "votes": null
    },
    {
      "id": "291884",
      "postDate": "03/07/2018 03:02:33",
      "content": "<p>Please check my <a href=\"https://www.kaggle.com/edzhang/time-gap-between-train-and-test/notebook\">kernel</a>.  I think it is true.</p>",
      "rawMarkdown": "Please check my [kernel][1].  I think it is true.\n  [1]: https://www.kaggle.com/edzhang/time-gap-between-train-and-test/notebook",
      "votes": null
    },
    {
      "id": "291905",
      "postDate": "03/07/2018 04:14:00",
      "content": "<p>Hi please check this as well: <a href=\"https://www.kaggle.com/yskn67/difference-between-train-and-test\">https://www.kaggle.com/yskn67/difference-between-train-and-test</a></p>\n\n<p>After i read that kernel I think what happend is the training data was 6-9th day of the month and the test data was 10th day of the month.</p>",
      "rawMarkdown": "Hi please check this as well: https://www.kaggle.com/yskn67/difference-between-train-and-test\n\nAfter i read that kernel I think what happend is the training data was 6-9th day of the month and the test data was 10th day of the month.",
      "votes": null
    },
    {
      "id": "291917",
      "postDate": "03/07/2018 04:44:38",
      "content": "<p>China is UTC+8 so if UTC is 1 AM China should be 9 AM. </p>",
      "rawMarkdown": "China is UTC+8 so if UTC is 1 AM China should be 9 AM.",
      "votes": null
    },
    {
      "id": "291918",
      "postDate": "03/07/2018 04:45:31",
      "content": "<p>To be a little more precise, the training data is 7-9th day of the month.</p>",
      "rawMarkdown": "To be a little more precise, the training data is 7-9th day of the month.",
      "votes": null
    },
    {
      "id": "291972",
      "postDate": "03/07/2018 06:47:19",
      "content": "<p>What I really care about is the 12 hours gap between 2017-11-10 04:00:00 and 2017-11-09 16:00:00. That will influence feature engineering a lot. </p>",
      "rawMarkdown": "What I really care about is the 12 hours gap between 2017-11-10 04:00:00 and 2017-11-09 16:00:00. That will influence feature engineering a lot.",
      "votes": null
    },
    {
      "id": "291976",
      "postDate": "03/07/2018 06:48:52",
      "content": "<p>thank you. </p>",
      "rawMarkdown": "thank you.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 291640,
      "author_name": "aaronyin",
      "author_url": "",
      "post_date": "03/06/2018 14:44:51",
      "content": "<p>I believe the timestamp is in UTC, not China local timezone. \nChina local timezone is 8 hours earlier than UTC, so it's actually 2017-11-09-23:59:59 in China local timezone. \nBut the min time in test set is a little bit wired, I will have a closer look.</p>",
      "votes": null,
      "replies": [
        {
          "id": 291847,
          "author_name": "edzhang",
          "author_url": "",
          "post_date": "03/07/2018 01:34:02",
          "content": "<p>\"But the min time in test set is a little bit wired\", That is right. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 291917,
          "author_name": "muhammadalfiansyah",
          "author_url": "",
          "post_date": "03/07/2018 04:44:38",
          "content": "<p>China is UTC+8 so if UTC is 1 AM China should be 9 AM. </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 291972,
          "author_name": "edzhang",
          "author_url": "",
          "post_date": "03/07/2018 06:47:19",
          "content": "<p>What I really care about is the 12 hours gap between 2017-11-10 04:00:00 and 2017-11-09 16:00:00. That will influence feature engineering a lot. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 291873,
      "author_name": "muhammadalfiansyah",
      "author_url": "",
      "post_date": "03/07/2018 02:38:16",
      "content": "<p>Is it true? if this true this data set might not be properly shuffled. </p>",
      "votes": null,
      "replies": [
        {
          "id": 291884,
          "author_name": "edzhang",
          "author_url": "",
          "post_date": "03/07/2018 03:02:33",
          "content": "<p>Please check my <a href=\"https://www.kaggle.com/edzhang/time-gap-between-train-and-test/notebook\">kernel</a>.  I think it is true.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 291905,
      "author_name": "muhammadalfiansyah",
      "author_url": "",
      "post_date": "03/07/2018 04:14:00",
      "content": "<p>Hi please check this as well: <a href=\"https://www.kaggle.com/yskn67/difference-between-train-and-test\">https://www.kaggle.com/yskn67/difference-between-train-and-test</a></p>\n\n<p>After i read that kernel I think what happend is the training data was 6-9th day of the month and the test data was 10th day of the month.</p>",
      "votes": null,
      "replies": [
        {
          "id": 291918,
          "author_name": "aaronyin",
          "author_url": "",
          "post_date": "03/07/2018 04:45:31",
          "content": "<p>To be a little more precise, the training data is 7-9th day of the month.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 291976,
          "author_name": "edzhang",
          "author_url": "",
          "post_date": "03/07/2018 06:48:52",
          "content": "<p>thank you. </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "291557": "Hello, the max timestamp of training set is 2017-11-09 16:00:00 , But the min timestamp of test set is 2017-11-10 04:00:00.\nI donot know why there is a gap between training set and test set. It is really bad to have this kind of gap.\nAnyone knows? Thank you in advance.",
    "291640": "I believe the timestamp is in UTC, not China local timezone. \nChina local timezone is 8 hours earlier than UTC, so it's actually 2017-11-09-23:59:59 in China local timezone. \nBut the min time in test set is a little bit wired, I will have a closer look.",
    "291847": "\"But the min time in test set is a little bit wired\", That is right.",
    "291873": "Is it true? if this true this data set might not be properly shuffled.",
    "291884": "Please check my [kernel][1].  I think it is true.\n  [1]: https://www.kaggle.com/edzhang/time-gap-between-train-and-test/notebook",
    "291905": "Hi please check this as well: https://www.kaggle.com/yskn67/difference-between-train-and-test\n\nAfter i read that kernel I think what happend is the training data was 6-9th day of the month and the test data was 10th day of the month.",
    "291917": "China is UTC+8 so if UTC is 1 AM China should be 9 AM.",
    "291918": "To be a little more precise, the training data is 7-9th day of the month.",
    "291972": "What I really care about is the 12 hours gap between 2017-11-10 04:00:00 and 2017-11-09 16:00:00. That will influence feature engineering a lot.",
    "291976": "thank you."
  },
  "source": "meta"
}