{
  "id": 154501,
  "title": "Metric will change?",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/154501",
  "author_name": "",
  "post_date": "2020-05-28T16:02:04.081560Z",
  "votes": 6,
  "comment_count": 3,
  "views": 0,
  "content": "<p>For some reason, I get a feeling that the evaluation metric may need to be changed. One day into the competition and already in <code>0.9xx</code> range with simple baselines. </p>",
  "messages": [
    {
      "id": "865415",
      "postDate": "05/28/2020 16:02:04",
      "content": "<p>For some reason, I get a feeling that the evaluation metric may need to be changed. One day into the competition and already in <code>0.9xx</code> range with simple baselines. </p>",
      "rawMarkdown": "For some reason, I get a feeling that the evaluation metric may need to be changed. One day into the competition and already in `0.9xx` range with simple baselines.",
      "votes": null
    },
    {
      "id": "865482",
      "postDate": "05/28/2020 16:47:22",
      "content": "<p>Damn, same feelings 😅 </p>",
      "rawMarkdown": "Damn, same feelings 😅",
      "votes": null
    },
    {
      "id": "865866",
      "postDate": "05/29/2020 00:10:02",
      "content": "<p>Not necessarily. Most of the \"Toxic\" NLP competitions ended up with AUC well in the 0.9x territory. For the first one you could get an average AUC of 0.98 with just a logistic regression.</p>",
      "rawMarkdown": "Not necessarily. Most of the \"Toxic\" NLP competitions ended up with AUC well in the 0.9x territory. For the first one you could get an average AUC of 0.98 with just a logistic regression.",
      "votes": null
    },
    {
      "id": "869909",
      "postDate": "06/01/2020 11:08:25",
      "content": "<p>I mean precision-recall curve might be useful here since we do not want false negatives.</p>",
      "rawMarkdown": "I mean precision-recall curve might be useful here since we do not want false negatives.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 865482,
      "author_name": "ipythonx",
      "author_url": "",
      "post_date": "05/28/2020 16:47:22",
      "content": "<p>Damn, same feelings 😅 </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 865866,
      "author_name": "tunguz",
      "author_url": "",
      "post_date": "05/29/2020 00:10:02",
      "content": "<p>Not necessarily. Most of the \"Toxic\" NLP competitions ended up with AUC well in the 0.9x territory. For the first one you could get an average AUC of 0.98 with just a logistic regression.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 869909,
      "author_name": "reighns",
      "author_url": "",
      "post_date": "06/01/2020 11:08:25",
      "content": "<p>I mean precision-recall curve might be useful here since we do not want false negatives.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "865415": "For some reason, I get a feeling that the evaluation metric may need to be changed. One day into the competition and already in `0.9xx` range with simple baselines.",
    "865482": "Damn, same feelings 😅",
    "865866": "Not necessarily. Most of the \"Toxic\" NLP competitions ended up with AUC well in the 0.9x territory. For the first one you could get an average AUC of 0.98 with just a logistic regression.",
    "869909": "I mean precision-recall curve might be useful here since we do not want false negatives."
  },
  "source": "meta"
}