{
  "id": 51942,
  "title": "How to deal with hour component of timestamp?",
  "url": "/competitions/talkingdata-adtracking-fraud-detection/discussion/51942",
  "author_name": "",
  "post_date": "2018-03-14T15:33:32.043463100Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>It is already clear that there is no reason to consider month and year.</p>\n\n<p>However, the hour of the day component might be somehow useful.</p>\n\n<p>Considering the time is in UTC timezone, do you think we should convert them all in UTC+8 because all the data are exclusively related to China?</p>",
  "messages": [
    {
      "id": "296011",
      "postDate": "03/14/2018 15:33:32",
      "content": "<p>It is already clear that there is no reason to consider month and year.</p>\n\n<p>However, the hour of the day component might be somehow useful.</p>\n\n<p>Considering the time is in UTC timezone, do you think we should convert them all in UTC+8 because all the data are exclusively related to China?</p>",
      "rawMarkdown": "It is already clear that there is no reason to consider month and year.\n\nHowever, the hour of the day component might be somehow useful.\n\nConsidering the time is in UTC timezone, do you think we should convert them all in UTC+8 because all the data are exclusively related to China?",
      "votes": null
    },
    {
      "id": "296066",
      "postDate": "03/14/2018 17:10:53",
      "content": "<p>I think converting can be useful for you to understand/interpret the data.  But as far as the models go, it won't make a difference, as long as all times in test and train are on the same timezone.</p>",
      "rawMarkdown": "I think converting can be useful for you to understand/interpret the data.  But as far as the models go, it won't make a difference, as long as all times in test and train are on the same timezone.",
      "votes": null
    },
    {
      "id": "296432",
      "postDate": "03/15/2018 07:49:45",
      "content": "<p>As yulia said. None of the models make a difference if the time is shifted with 8 hours. Just make sure that if you do change the timezone you change it for the whole dataset (train+test). Given that you can go wrong by forgetting to change something, I will just leave the hour as it is.</p>",
      "rawMarkdown": "As yulia said. None of the models make a difference if the time is shifted with 8 hours. Just make sure that if you do change the timezone you change it for the whole dataset (train+test). Given that you can go wrong by forgetting to change something, I will just leave the hour as it is.",
      "votes": null
    },
    {
      "id": "296470",
      "postDate": "03/15/2018 09:17:18",
      "content": "<p>I don't think there's a need to convert anything - all the action 'happens' in the same time zone, and that's all that matter imo.</p>",
      "rawMarkdown": "I don't think there's a need to convert anything - all the action 'happens' in the same time zone, and that's all that matter imo.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 296066,
      "author_name": "yuliagm",
      "author_url": "",
      "post_date": "03/14/2018 17:10:53",
      "content": "<p>I think converting can be useful for you to understand/interpret the data.  But as far as the models go, it won't make a difference, as long as all times in test and train are on the same timezone.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 296432,
      "author_name": "asparuhhristov",
      "author_url": "",
      "post_date": "03/15/2018 07:49:45",
      "content": "<p>As yulia said. None of the models make a difference if the time is shifted with 8 hours. Just make sure that if you do change the timezone you change it for the whole dataset (train+test). Given that you can go wrong by forgetting to change something, I will just leave the hour as it is.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 296470,
      "author_name": "konradb",
      "author_url": "",
      "post_date": "03/15/2018 09:17:18",
      "content": "<p>I don't think there's a need to convert anything - all the action 'happens' in the same time zone, and that's all that matter imo.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "296011": "It is already clear that there is no reason to consider month and year.\n\nHowever, the hour of the day component might be somehow useful.\n\nConsidering the time is in UTC timezone, do you think we should convert them all in UTC+8 because all the data are exclusively related to China?",
    "296066": "I think converting can be useful for you to understand/interpret the data.  But as far as the models go, it won't make a difference, as long as all times in test and train are on the same timezone.",
    "296432": "As yulia said. None of the models make a difference if the time is shifted with 8 hours. Just make sure that if you do change the timezone you change it for the whole dataset (train+test). Given that you can go wrong by forgetting to change something, I will just leave the hour as it is.",
    "296470": "I don't think there's a need to convert anything - all the action 'happens' in the same time zone, and that's all that matter imo."
  },
  "source": "meta"
}