{
  "id": 1535,
  "title": "Evaluation Formula",
  "url": "/competitions/benchmark-bond-trade-price-challenge/discussion/1535",
  "author_name": "",
  "post_date": "2012-03-14T12:35:05.397Z",
  "votes": null,
  "comment_count": 3,
  "views": 3548,
  "content": "<p>Can anyone please explain the exact formula used for evaluation?</p>",
  "messages": [
    {
      "id": "9305",
      "postDate": "03/14/2012 12:35:05",
      "content": "<p>Can anyone please explain the exact formula used for evaluation?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "9317",
      "postDate": "03/14/2012 20:20:05",
      "content": "<p>I'm pretty sure it's just the average of weight * abs(trade_price - prediction).</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "10292",
      "postDate": "04/23/2012 06:07:27",
      "content": "<p>2Glen</p>\r\n<p>Are you really sure? So then I must think that we don't have this &quot;weight&quot;? Because when I try to get mean absolute error of my data without weight (from test array) I obtain error less than 0.7 but if I use your formula with weight * I get mean absolute\r\n error more than 0.7.</p>\r\n<p>And my result now on leaderboard&nbsp;<a title=\"1w improvement: none\" href=\"http://www.kaggle.com/c/benchmark-bond-trade-price-challenge/leaderboard?submissionId=111882#1.14569\">1.14569</a>&nbsp;how can it be?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "10299",
      "postDate": "04/23/2012 13:24:30",
      "content": "<p>The above response is correct. &nbsp;You can read the &quot;Evaluation&quot; tab for a slightly more complete answer.</p>\r\n<p>-Dan</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 9317,
      "author_name": "gkoundry",
      "author_url": "",
      "post_date": "03/14/2012 20:20:05",
      "content": "<p>I'm pretty sure it's just the average of weight * abs(trade_price - prediction).</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 10292,
      "author_name": "skyline",
      "author_url": "",
      "post_date": "04/23/2012 06:07:27",
      "content": "<p>2Glen</p>\r\n<p>Are you really sure? So then I must think that we don't have this &quot;weight&quot;? Because when I try to get mean absolute error of my data without weight (from test array) I obtain error less than 0.7 but if I use your formula with weight * I get mean absolute\r\n error more than 0.7.</p>\r\n<p>And my result now on leaderboard&nbsp;<a title=\"1w improvement: none\" href=\"http://www.kaggle.com/c/benchmark-bond-trade-price-challenge/leaderboard?submissionId=111882#1.14569\">1.14569</a>&nbsp;how can it be?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 10299,
      "author_name": "danglaser",
      "author_url": "",
      "post_date": "04/23/2012 13:24:30",
      "content": "<p>The above response is correct. &nbsp;You can read the &quot;Evaluation&quot; tab for a slightly more complete answer.</p>\r\n<p>-Dan</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "9305": "",
    "9317": "",
    "10292": "",
    "10299": ""
  },
  "source": "meta"
}