{
  "id": 10964,
  "title": "No country field on new dataset?",
  "url": "/competitions/avazu-ctr-prediction/discussion/10964",
  "author_name": "",
  "post_date": "2014-11-19T11:28:04.323Z",
  "votes": null,
  "comment_count": 2,
  "views": 1213,
  "content": "<p>It looks like there is no country field on the new dataset (or am I missing something)? &nbsp;As others have mentioned, on the old dataset it was useful to cross country with hour as a proxy for timezone. &nbsp;I'm trying to think up some new timezone proxies, does anyone have any ideas?</p>",
  "messages": [
    {
      "id": "58391",
      "postDate": "11/19/2014 11:28:04",
      "content": "<p>It looks like there is no country field on the new dataset (or am I missing something)? &nbsp;As others have mentioned, on the old dataset it was useful to cross country with hour as a proxy for timezone. &nbsp;I'm trying to think up some new timezone proxies, does anyone have any ideas?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "58667",
      "postDate": "11/22/2014 04:38:09",
      "content": "<p>Yes, it seems that there is no country in the new dataset and thus, timezone which could have been a very predictive variable is now gone.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "58711",
      "postDate": "11/23/2014 02:04:16",
      "content": "<p>I actually read we're not supposed to use the old data, so I deleted it. &nbsp;You're thoughts about the C* variables are interesting though. &nbsp;I worry that it would tough to tease out any relationship between anonymized lat/long variables, but if there were some other more categorical geographical/longitudinal information encoded in one of the C* variables you might be able to pick it out visually by just graphing the frequency of observations for the top few values across the hours of the day, I may give that a try.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 58667,
      "author_name": "nishkgp",
      "author_url": "",
      "post_date": "11/22/2014 04:38:09",
      "content": "<p>Yes, it seems that there is no country in the new dataset and thus, timezone which could have been a very predictive variable is now gone.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 58711,
      "author_name": "dlsiii",
      "author_url": "",
      "post_date": "11/23/2014 02:04:16",
      "content": "<p>I actually read we're not supposed to use the old data, so I deleted it. &nbsp;You're thoughts about the C* variables are interesting though. &nbsp;I worry that it would tough to tease out any relationship between anonymized lat/long variables, but if there were some other more categorical geographical/longitudinal information encoded in one of the C* variables you might be able to pick it out visually by just graphing the frequency of observations for the top few values across the hours of the day, I may give that a try.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "58391": "",
    "58667": "",
    "58711": ""
  },
  "source": "meta"
}