{
  "id": 228736,
  "title": "Be confused about f1 score",
  "url": "/competitions/plant-pathology-2021-fgvc8/discussion/228736",
  "author_name": "",
  "post_date": "2021-03-26T06:45:16.792031200Z",
  "votes": 4,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I'm confused about the definition of f1 score. What does ' Mean F1-Score' in the overview refer to? 'macro', 'micro', or 'weighted' ? ‘macro’ is used in my cvs, but there is a big gap between cv and LB.</p>",
  "messages": [
    {
      "id": "1252893",
      "postDate": "03/26/2021 06:45:16",
      "content": "<p>I'm confused about the definition of f1 score. What does ' Mean F1-Score' in the overview refer to? 'macro', 'micro', or 'weighted' ? ‘macro’ is used in my cvs, but there is a big gap between cv and LB.</p>",
      "rawMarkdown": "I'm confused about the definition of f1 score. What does ' Mean F1-Score' in the overview refer to? 'macro', 'micro', or 'weighted' ? ‘macro’ is used in my cvs, but there is a big gap between cv and LB.",
      "votes": null
    },
    {
      "id": "1252918",
      "postDate": "03/26/2021 07:28:31",
      "content": "<p>Hello!</p>\n<p>I've opened  discussions both on <strong><a href=\"https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/228422\" target=\"_blank\">Unclarified metric</a></strong> and <strong><a href=\"https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/227237\" target=\"_blank\">CV-LB inconsistency</a></strong> a while ago. There's still no information from the organizers for the former, and for the latter, it's still unclear.</p>\n<p>My recent CV F1 score is <code>0.9091</code> with <code>average=macro</code> (and over 97% accuracy), but I hardly manage to get 0.6XX LB.</p>",
      "rawMarkdown": "Hello!\n\nI've opened  discussions both on **[Unclarified metric](https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/228422)** and **[CV-LB inconsistency](https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/227237)** a while ago. There's still no information from the organizers for the former, and for the latter, it's still unclear.\n\nMy recent CV F1 score is `0.9091` with `average=macro` (and over 97% accuracy), but I hardly manage to get 0.6XX LB.",
      "votes": null
    },
    {
      "id": "1252929",
      "postDate": "03/26/2021 07:42:00",
      "content": "<p><a href=\"https://www.kaggle.com/fruitpathology\" target=\"_blank\">@fruitpathology</a>  we need more information…😬</p>",
      "rawMarkdown": "fruitpathology  we need more information...😬",
      "votes": null
    },
    {
      "id": "1264109",
      "postDate": "04/05/2021 22:01:11",
      "content": "<p>It's closest to sklearn's <code>average='samples'</code>, though there might be some small differences in our implementations.</p>",
      "rawMarkdown": "It's closest to sklearn's `average='samples'`, though there might be some small differences in our implementations.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1252918,
      "author_name": "nickuzmenkov",
      "author_url": "",
      "post_date": "03/26/2021 07:28:31",
      "content": "<p>Hello!</p>\n<p>I've opened  discussions both on <strong><a href=\"https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/228422\" target=\"_blank\">Unclarified metric</a></strong> and <strong><a href=\"https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/227237\" target=\"_blank\">CV-LB inconsistency</a></strong> a while ago. There's still no information from the organizers for the former, and for the latter, it's still unclear.</p>\n<p>My recent CV F1 score is <code>0.9091</code> with <code>average=macro</code> (and over 97% accuracy), but I hardly manage to get 0.6XX LB.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1252929,
          "author_name": "eynaij",
          "author_url": "",
          "post_date": "03/26/2021 07:42:00",
          "content": "<p><a href=\"https://www.kaggle.com/fruitpathology\" target=\"_blank\">@fruitpathology</a>  we need more information…😬</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1264109,
      "author_name": "sohier",
      "author_url": "",
      "post_date": "04/05/2021 22:01:11",
      "content": "<p>It's closest to sklearn's <code>average='samples'</code>, though there might be some small differences in our implementations.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1252893": "I'm confused about the definition of f1 score. What does ' Mean F1-Score' in the overview refer to? 'macro', 'micro', or 'weighted' ? ‘macro’ is used in my cvs, but there is a big gap between cv and LB.",
    "1252918": "Hello!\n\nI've opened  discussions both on **[Unclarified metric](https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/228422)** and **[CV-LB inconsistency](https://www.kaggle.com/c/plant-pathology-2021-fgvc8/discussion/227237)** a while ago. There's still no information from the organizers for the former, and for the latter, it's still unclear.\n\nMy recent CV F1 score is `0.9091` with `average=macro` (and over 97% accuracy), but I hardly manage to get 0.6XX LB.",
    "1252929": "fruitpathology  we need more information...😬",
    "1264109": "It's closest to sklearn's `average='samples'`, though there might be some small differences in our implementations."
  },
  "source": "meta"
}