{
  "id": 5626,
  "title": "Purpose with Accuracy So High",
  "url": "/competitions/belkin-energy-disaggregation-competition/discussion/5626",
  "author_name": "",
  "post_date": "2013-09-02T20:18:07.707Z",
  "votes": 1,
  "comment_count": 4,
  "views": 1595,
  "content": "<p>I just began looking at this competition today and familiarizing myself with the material. Reading through the slide deck, if the accuracy is already so high and devices can be identified with incredible accuracy, what is this competition trying to accomplish? Or am I missing something?</p>",
  "messages": [
    {
      "id": "30008",
      "postDate": "09/02/2013 20:18:07",
      "content": "<p>I just began looking at this competition today and familiarizing myself with the material. Reading through the slide deck, if the accuracy is already so high and devices can be identified with incredible accuracy, what is this competition trying to accomplish? Or am I missing something?</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30052",
      "postDate": "09/03/2013 13:51:08",
      "content": "<p>The accuracy now is not high in practical sence, because 0.07877 is &quot;zero&quot; level.</p>\n<p>But to improve it one need to make prediction with posterior (conditional) probability &gt;0.5 when prior one is 0.07877. This indeed looks to be incredible accuracy.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30075",
      "postDate": "09/03/2013 18:59:55",
      "content": "<p>He's talking about the papers and the slide deck. In reference [4] on the main page, a bunch of 10-fold cv accuracies are quoted ranging from 89-100%. &nbsp;</p>\n<p>I would like to know the answer to this question, too.</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30081",
      "postDate": "09/03/2013 20:04:23",
      "content": "<p>Yeah, this was niggling at me as well...</p>",
      "rawMarkdown": "",
      "votes": null
    },
    {
      "id": "30173",
      "postDate": "09/04/2013 17:16:53",
      "content": "<p>In reference [4] on the main page there is 93.82% accuracy reported, that is only slightly better than (92.12%=1-0.07877) - the accuracy of zero benchmark.</p>",
      "rawMarkdown": "",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 30052,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "09/03/2013 13:51:08",
      "content": "<p>The accuracy now is not high in practical sence, because 0.07877 is &quot;zero&quot; level.</p>\n<p>But to improve it one need to make prediction with posterior (conditional) probability &gt;0.5 when prior one is 0.07877. This indeed looks to be incredible accuracy.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30075,
      "author_name": "davidthaler",
      "author_url": "",
      "post_date": "09/03/2013 18:59:55",
      "content": "<p>He's talking about the papers and the slide deck. In reference [4] on the main page, a bunch of 10-fold cv accuracies are quoted ranging from 89-100%. &nbsp;</p>\n<p>I would like to know the answer to this question, too.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30081,
      "author_name": "nitaidean",
      "author_url": "",
      "post_date": "09/03/2013 20:04:23",
      "content": "<p>Yeah, this was niggling at me as well...</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 30173,
      "author_name": "nedelko",
      "author_url": "",
      "post_date": "09/04/2013 17:16:53",
      "content": "<p>In reference [4] on the main page there is 93.82% accuracy reported, that is only slightly better than (92.12%=1-0.07877) - the accuracy of zero benchmark.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "30008": "",
    "30052": "",
    "30075": "",
    "30081": "",
    "30173": ""
  },
  "source": "meta"
}