{
  "id": 213802,
  "title": "About the class of 'CVC-borderline'!",
  "url": "/competitions/ranzcr-clip-catheter-line-classification/discussion/213802",
  "author_name": "",
  "post_date": "2021-01-24T11:15:08.650617100Z",
  "votes": 5,
  "comment_count": 4,
  "views": 0,
  "content": "<p>In my predicted result,I find the 'CVC-boderline' class auc score is the lowest,and it can't reach 0.90,it has a big difference with other class.Anyone cares about this problem?Thanks!</p>",
  "messages": [
    {
      "id": "1167576",
      "postDate": "01/24/2021 11:15:08",
      "content": "<p>In my predicted result,I find the 'CVC-boderline' class auc score is the lowest,and it can't reach 0.90,it has a big difference with other class.Anyone cares about this problem?Thanks!</p>",
      "rawMarkdown": "In my predicted result,I find the 'CVC-boderline' class auc score is the lowest,and it can't reach 0.90,it has a big difference with other class.Anyone cares about this problem?Thanks!",
      "votes": null
    },
    {
      "id": "1171127",
      "postDate": "01/26/2021 16:33:33",
      "content": "<p><a href=\"https://www.kaggle.com/bcwang\" target=\"_blank\">@bcwang</a> I also faced the same problem issue, I even tried to make a different model for CVC class but faced the same issue with CVC borderline.</p>",
      "rawMarkdown": "bcwang I also faced the same problem issue, I even tried to make a different model for CVC class but faced the same issue with CVC borderline.",
      "votes": null
    },
    {
      "id": "1171564",
      "postDate": "01/27/2021 00:25:40",
      "content": "<p>I've seen it discussed elsewhere by raddar and hengck, but the problem comes down to the wide spatial distribution for the CVC borderline cases, plus some interobserver variability. As an example of the former, a CVC may be borderline if it is above the right atrium, below the right atrium, or <em>anywhere else along the trajectory of the CVC</em>, which may be the periphery (arm or neck). Hope that helps! </p>",
      "rawMarkdown": "I've seen it discussed elsewhere by raddar and hengck, but the problem comes down to the wide spatial distribution for the CVC borderline cases, plus some interobserver variability. As an example of the former, a CVC may be borderline if it is above the right atrium, below the right atrium, or _anywhere else along the trajectory of the CVC_, which may be the periphery (arm or neck). Hope that helps!",
      "votes": null
    },
    {
      "id": "1171601",
      "postDate": "01/27/2021 01:31:47",
      "content": "<p>Thanks for you reply!This is give me a lot help!</p>",
      "rawMarkdown": "Thanks for you reply!This is give me a lot help!",
      "votes": null
    },
    {
      "id": "1183176",
      "postDate": "02/02/2021 19:41:13",
      "content": "<p>Because there are so many of them. AUC is hard to optimize in this case.</p>",
      "rawMarkdown": "Because there are so many of them. AUC is hard to optimize in this case.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1171127,
      "author_name": "agsam23",
      "author_url": "",
      "post_date": "01/26/2021 16:33:33",
      "content": "<p><a href=\"https://www.kaggle.com/bcwang\" target=\"_blank\">@bcwang</a> I also faced the same problem issue, I even tried to make a different model for CVC class but faced the same issue with CVC borderline.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1171564,
      "author_name": "reubenschmidt",
      "author_url": "",
      "post_date": "01/27/2021 00:25:40",
      "content": "<p>I've seen it discussed elsewhere by raddar and hengck, but the problem comes down to the wide spatial distribution for the CVC borderline cases, plus some interobserver variability. As an example of the former, a CVC may be borderline if it is above the right atrium, below the right atrium, or <em>anywhere else along the trajectory of the CVC</em>, which may be the periphery (arm or neck). Hope that helps! </p>",
      "votes": null,
      "replies": [
        {
          "id": 1171601,
          "author_name": "bcwang",
          "author_url": "",
          "post_date": "01/27/2021 01:31:47",
          "content": "<p>Thanks for you reply!This is give me a lot help!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1183176,
      "author_name": "underwearfitting",
      "author_url": "",
      "post_date": "02/02/2021 19:41:13",
      "content": "<p>Because there are so many of them. AUC is hard to optimize in this case.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1167576": "In my predicted result,I find the 'CVC-boderline' class auc score is the lowest,and it can't reach 0.90,it has a big difference with other class.Anyone cares about this problem?Thanks!",
    "1171127": "bcwang I also faced the same problem issue, I even tried to make a different model for CVC class but faced the same issue with CVC borderline.",
    "1171564": "I've seen it discussed elsewhere by raddar and hengck, but the problem comes down to the wide spatial distribution for the CVC borderline cases, plus some interobserver variability. As an example of the former, a CVC may be borderline if it is above the right atrium, below the right atrium, or _anywhere else along the trajectory of the CVC_, which may be the periphery (arm or neck). Hope that helps!",
    "1171601": "Thanks for you reply!This is give me a lot help!",
    "1183176": "Because there are so many of them. AUC is hard to optimize in this case."
  },
  "source": "meta"
}