{
  "id": 5725,
  "title": "I need professional help",
  "url": "/competitions/expedia-personalized-sort/discussion/5725",
  "author_name": "",
  "post_date": "2013-09-11T15:45:24.163Z",
  "votes": null,
  "comment_count": 11,
  "views": 3482,
  "content": "<p>My score is .04, I should get an award for the worst score ever. &nbsp;In all seriousness, if anyone has a guess of why my rankings are going so wrong I would greatly appreciate it. &nbsp;Does the evaluator check to make sure the solution is &quot;valid&quot;, i.e. every id is represented once and only once?</p>\n<p>I've tried to do all sorts of validation on the training data and my NDCG values match the random sort benchmark value when I use random rankings, and in general my NDCG scores on the training data are reasonable compared to everyone else's on the leader board. &nbsp;If I didn't know better, I would think I was doing an ascending sort instead of descending.</p>\n<p>&nbsp;</p>",
  "messages": [
    {
      "id": "30734",
      "postDate": "09/11/2013 15:45:24",
      "content": "<p>My score is .04, I should get an award for the worst score ever. &nbsp;In all seriousness, if anyone has a guess of why my rankings are going so wrong I would greatly appreciate it. &nbsp;Does the evaluator check to make sure the solution is &quot;valid&quot;, i.e. every id is represented once and only once?</p>\n<p>I've tried to do all sorts of validation on the training data and my NDCG values match the random sort benchmark value when I use random rankings, and in general my NDCG scores on the training data are reasonable compared to everyone else's on the leader board. &nbsp;If I didn't know better, I would think I was doing an ascending sort instead of descending.</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30736",
      "postDate": "09/11/2013 16:21:14",
      "content": "<p>If you post your first rows (100) of your sent I'll try help you if there is a trivial bug.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30739",
      "postDate": "09/11/2013 16:30:58",
      "content": "<p>That's very kind of you.</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30740",
      "postDate": "09/11/2013 16:41:06",
      "content": "<p>If you change the scientific format for property_id, your score will be good as I see.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30743",
      "postDate": "09/11/2013 17:40:01",
      "content": "<p>Thanks for your help, that certainly may have been part of the issue. &nbsp;I tried to fix it (in Matlab) using this</p>\n<p>for i=1:N<br>fprintf(fid,'%.f, %.f\\n',solution(i,1),solution(i,2));<br>end</p>\n<p>and the scientific notation is gone, but I now get a score of zero!</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30744",
      "postDate": "09/11/2013 17:43:34",
      "content": "<p>post the new 100 first rows again</p>\n<p>&nbsp;</p>\n<p>EDIT, I think your problem may be the thousand separator. Try export the property_id to character</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30745",
      "postDate": "09/11/2013 17:47:26",
      "content": "<p>Okay, here is the first rows of the latest attempt</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30748",
      "postDate": "09/11/2013 17:56:55",
      "content": "<p>The format is OK and the predictions seems good.</p>\n<p>Sincerely I don't know why score zero.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30749",
      "postDate": "09/11/2013 18:01:23",
      "content": "<p>Thanks for trying!</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30765",
      "postDate": "09/12/2013 01:25:53",
      "content": "<p>if the problem is format issue, will the evaluation to be 0?</p>\n<p>so i guess the 2nd attach file format is fine, correct?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30766",
      "postDate": "09/12/2013 01:36:22",
      "content": "<p>That's what I'm confused about. &nbsp;If my file format is correct, then why did I get a score of zero? Clearly the first few id's should have given me some small positive score. &nbsp;Unfortunately with the 2 entries per day I'm having trouble figuring this one out.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30767",
      "postDate": "09/12/2013 01:59:10",
      "content": "<p>It worked! &nbsp;I have no idea why, which is kind of unsettling. &nbsp;Thanks again, Jose, for looking at my .csv files. &nbsp;</p>\n<p>&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 30736,
      "author_name": "blindape",
      "author_url": "",
      "post_date": "09/11/2013 16:21:14",
      "content": "<p>If you post your first rows (100) of your sent I'll try help you if there is a trivial bug.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30739,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/11/2013 16:30:58",
      "content": "<p>That's very kind of you.</p>\n<p>&nbsp;</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30740,
      "author_name": "blindape",
      "author_url": "",
      "post_date": "09/11/2013 16:41:06",
      "content": "<p>If you change the scientific format for property_id, your score will be good as I see.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30743,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/11/2013 17:40:01",
      "content": "<p>Thanks for your help, that certainly may have been part of the issue. &nbsp;I tried to fix it (in Matlab) using this</p>\n<p>for i=1:N<br>fprintf(fid,'%.f, %.f\\n',solution(i,1),solution(i,2));<br>end</p>\n<p>and the scientific notation is gone, but I now get a score of zero!</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30744,
      "author_name": "blindape",
      "author_url": "",
      "post_date": "09/11/2013 17:43:34",
      "content": "<p>post the new 100 first rows again</p>\n<p>&nbsp;</p>\n<p>EDIT, I think your problem may be the thousand separator. Try export the property_id to character</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30745,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/11/2013 17:47:26",
      "content": "<p>Okay, here is the first rows of the latest attempt</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30748,
      "author_name": "blindape",
      "author_url": "",
      "post_date": "09/11/2013 17:56:55",
      "content": "<p>The format is OK and the predictions seems good.</p>\n<p>Sincerely I don't know why score zero.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30749,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/11/2013 18:01:23",
      "content": "<p>Thanks for trying!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30765,
      "author_name": "zhaogang",
      "author_url": "",
      "post_date": "09/12/2013 01:25:53",
      "content": "<p>if the problem is format issue, will the evaluation to be 0?</p>\n<p>so i guess the 2nd attach file format is fine, correct?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30766,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/12/2013 01:36:22",
      "content": "<p>That's what I'm confused about. &nbsp;If my file format is correct, then why did I get a score of zero? Clearly the first few id's should have given me some small positive score. &nbsp;Unfortunately with the 2 entries per day I'm having trouble figuring this one out.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30767,
      "author_name": "george04",
      "author_url": "",
      "post_date": "09/12/2013 01:59:10",
      "content": "<p>It worked! &nbsp;I have no idea why, which is kind of unsettling. &nbsp;Thanks again, Jose, for looking at my .csv files. &nbsp;</p>\n<p>&nbsp;</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "30734": "",
    "30736": "",
    "30739": "",
    "30740": "",
    "30743": "",
    "30744": "",
    "30745": "",
    "30748": "",
    "30749": "",
    "30765": "",
    "30766": "",
    "30767": ""
  },
  "source": "meta"
}