{
  "id": 6450,
  "title": "Must one submit at least one prediction for each post?",
  "url": "/competitions/facebook-recruiting-iii-keyword-extraction/discussion/6450",
  "author_name": "",
  "post_date": "2013-11-26T18:55:16.707Z",
  "votes": null,
  "comment_count": 5,
  "views": 1534,
  "content": "<p>And if not...can I ask for guidance on the accepted method/format of submitting no predictions for a post in our submission file?</p>",
  "messages": [
    {
      "id": "35304",
      "postDate": "11/26/2013 18:55:16",
      "content": "<p>And if not...can I ask for guidance on the accepted method/format of submitting no predictions for a post in our submission file?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "35306",
      "postDate": "11/26/2013 19:12:48",
      "content": "<p>Some of my observations had no prediction. The Tags value just looked like &quot;&quot;.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "35307",
      "postDate": "11/26/2013 19:20:12",
      "content": "<p>Why do you want to submit your file with no predictions for some posts?</p>\n<p>In a post, the F1 score with no predictions and when your predictions are wrong is 0. A random prediction can only improve your score.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "35309",
      "postDate": "11/26/2013 19:32:00",
      "content": "<p>I'm &nbsp;a bit tired after my last thread, so I'm gonna let people believe what they want to believe. I'm still interested in the answer, and am assuming the method is to put &quot;&quot; as the prediction for now.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "35311",
      "postDate": "11/26/2013 19:52:55",
      "content": "<p>In the solution format says that your need a line for every post of the test set. If &quot;&quot; doesn't work you can use &quot;MyInventedTag&quot; as a prediction.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "35428",
      "postDate": "11/28/2013 08:39:41",
      "content": "<p>gah. &nbsp;My bad. &nbsp;I've been working under the notion that it was the F1 score for the entire data set that we were competing to maximize, whereas its the individual F1 score for each row which is then averaged...two deceptively different things...</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 35306,
      "author_name": "adjgiulio",
      "author_url": "",
      "post_date": "11/26/2013 19:12:48",
      "content": "<p>Some of my observations had no prediction. The Tags value just looked like &quot;&quot;.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 35307,
      "author_name": "robertod",
      "author_url": "",
      "post_date": "11/26/2013 19:20:12",
      "content": "<p>Why do you want to submit your file with no predictions for some posts?</p>\n<p>In a post, the F1 score with no predictions and when your predictions are wrong is 0. A random prediction can only improve your score.</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 35309,
      "author_name": "damienmelksham",
      "author_url": "",
      "post_date": "11/26/2013 19:32:00",
      "content": "<p>I'm &nbsp;a bit tired after my last thread, so I'm gonna let people believe what they want to believe. I'm still interested in the answer, and am assuming the method is to put &quot;&quot; as the prediction for now.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 35311,
      "author_name": "robertod",
      "author_url": "",
      "post_date": "11/26/2013 19:52:55",
      "content": "<p>In the solution format says that your need a line for every post of the test set. If &quot;&quot; doesn't work you can use &quot;MyInventedTag&quot; as a prediction.</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 35428,
      "author_name": "damienmelksham",
      "author_url": "",
      "post_date": "11/28/2013 08:39:41",
      "content": "<p>gah. &nbsp;My bad. &nbsp;I've been working under the notion that it was the F1 score for the entire data set that we were competing to maximize, whereas its the individual F1 score for each row which is then averaged...two deceptively different things...</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "35304": "",
    "35306": "",
    "35307": "",
    "35309": "",
    "35311": "",
    "35428": ""
  },
  "source": "meta"
}