{
  "id": 10648,
  "title": "How is the score calculated?",
  "url": "/competitions/seizure-prediction/discussion/10648",
  "author_name": "",
  "post_date": "2014-10-19T14:46:14.337Z",
  "votes": 2,
  "comment_count": 2,
  "views": 929,
  "content": "<p>I would like to understand how is the score for this challenge calculated. The description says the score is the AUC. Once someone uploads their results, let the true positive be TP and false positive be FP (this will only give one point on the curve). Then is the AUC and hence the score calculated as:</p>\n<p>1/2*TP*FP + (1-FP)*TP + 1/2*(1-TP)*(1-FP)</p>\n<p>or is there some other calculation (what?) used to calculate the score and AUC?</p>",
  "messages": [
    {
      "id": "56257",
      "postDate": "10/19/2014 14:46:14",
      "content": "<p>I would like to understand how is the score for this challenge calculated. The description says the score is the AUC. Once someone uploads their results, let the true positive be TP and false positive be FP (this will only give one point on the curve). Then is the AUC and hence the score calculated as:</p>\n<p>1/2*TP*FP + (1-FP)*TP + 1/2*(1-TP)*(1-FP)</p>\n<p>or is there some other calculation (what?) used to calculate the score and AUC?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "56259",
      "postDate": "10/19/2014 16:52:39",
      "content": "<p>This was also unclear to me from the problem description, but instead of just classes 0 or 1 you are allowed to submit floating point scores (between 0 and 1) which are then used to compute your AUC (see&nbsp;https://www.kaggle.com/wiki/AreaUnderCurve ). Doing so instead of only submitting 0's and 1's is also likely to improve your score.&nbsp;</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "56261",
      "postDate": "10/19/2014 17:17:34",
      "content": "<p>thanks for the link. i think that explains it.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 56259,
      "author_name": "oseiskar",
      "author_url": "",
      "post_date": "10/19/2014 16:52:39",
      "content": "<p>This was also unclear to me from the problem description, but instead of just classes 0 or 1 you are allowed to submit floating point scores (between 0 and 1) which are then used to compute your AUC (see&nbsp;https://www.kaggle.com/wiki/AreaUnderCurve ). Doing so instead of only submitting 0's and 1's is also likely to improve your score.&nbsp;</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 56261,
      "author_name": "siddjain",
      "author_url": "",
      "post_date": "10/19/2014 17:17:34",
      "content": "<p>thanks for the link. i think that explains it.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "56257": "",
    "56259": "",
    "56261": ""
  },
  "source": "meta"
}