{
  "id": 290136,
  "title": "Understanding the Metric : F2 score",
  "url": "/competitions/tensorflow-great-barrier-reef/discussion/290136",
  "author_name": "",
  "post_date": "2021-11-23T07:27:08.360787100Z",
  "votes": 43,
  "comment_count": 2,
  "views": 0,
  "content": "<p>We all have heard about the F1 score, which is just the harmonic mean of Precision and Recall<br>\n(If you want to revise recall and precision you can see here <a href=\"https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall\" target=\"_blank\">https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall</a>)<br>\nWhat the F1 score is doing is that it gives equal weightage to Recall and Precision. But sometimes according to our need, we may prefer one of Recall/Precision more than the other.<br>\nA generalised form of F score in terms of 'beta' can be<br>\n<code>F_beta =((1 + beta^2) * Precision * Recall) / (beta^2 * Precision + Recall)</code><br>\nwhere if beta = 1 it gives us the F1 score.<br>\nHere in this task, we are using <code>beta = 2,</code><br>\nhence <code>F2  score = (5 * Precision * Recall) / (4 * Precision + Recall)</code><br>\n F2-measure puts more weight on minimizing false negatives than minimizing false positives. It has the effect of increasing the importance of Recall.<br>\nYou can look more here <a href=\"https://machinelearningmastery.com/fbeta-measure-for-machine-learning/\" target=\"_blank\">https://machinelearningmastery.com/fbeta-measure-for-machine-learning/</a></p>",
  "messages": [
    {
      "id": "1592587",
      "postDate": "11/23/2021 07:27:08",
      "content": "<p>We all have heard about the F1 score, which is just the harmonic mean of Precision and Recall<br>\n(If you want to revise recall and precision you can see here <a href=\"https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall\" target=\"_blank\">https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall</a>)<br>\nWhat the F1 score is doing is that it gives equal weightage to Recall and Precision. But sometimes according to our need, we may prefer one of Recall/Precision more than the other.<br>\nA generalised form of F score in terms of 'beta' can be<br>\n<code>F_beta =((1 + beta^2) * Precision * Recall) / (beta^2 * Precision + Recall)</code><br>\nwhere if beta = 1 it gives us the F1 score.<br>\nHere in this task, we are using <code>beta = 2,</code><br>\nhence <code>F2  score = (5 * Precision * Recall) / (4 * Precision + Recall)</code><br>\n F2-measure puts more weight on minimizing false negatives than minimizing false positives. It has the effect of increasing the importance of Recall.<br>\nYou can look more here <a href=\"https://machinelearningmastery.com/fbeta-measure-for-machine-learning/\" target=\"_blank\">https://machinelearningmastery.com/fbeta-measure-for-machine-learning/</a></p>",
      "rawMarkdown": "We all have heard about the F1 score, which is just the harmonic mean of Precision and Recall\n(If you want to revise recall and precision you can see here https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall)\nWhat the F1 score is doing is that it gives equal weightage to Recall and Precision. But sometimes according to our need, we may prefer one of Recall/Precision more than the other.\nA generalised form of F score in terms of 'beta' can be\n`F_beta =((1 + beta^2) * Precision * Recall) / (beta^2 * Precision + Recall)`\nwhere if beta = 1 it gives us the F1 score.\nHere in this task, we are using `beta = 2,`\nhence `F2  score = (5 * Precision * Recall) / (4 * Precision + Recall) `\n F2-measure puts more weight on minimizing false negatives than minimizing false positives. It has the effect of increasing the importance of Recall.\nYou can look more here https://machinelearningmastery.com/fbeta-measure-for-machine-learning/",
      "votes": null
    },
    {
      "id": "1593313",
      "postDate": "11/23/2021 19:08:05",
      "content": "<p>here we are assuming the star fish as the positive and the other species are the negative example.<br>\nPlease correct me if I'm wrong.</p>",
      "rawMarkdown": "here we are assuming the star fish as the positive and the other species are the negative example.\nPlease correct me if I'm wrong.",
      "votes": null
    },
    {
      "id": "1593338",
      "postDate": "11/23/2021 19:38:09",
      "content": "<p>yes here we are doing that</p>",
      "rawMarkdown": "yes here we are doing that",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1593313,
      "author_name": "harshitlakhani",
      "author_url": "",
      "post_date": "11/23/2021 19:08:05",
      "content": "<p>here we are assuming the star fish as the positive and the other species are the negative example.<br>\nPlease correct me if I'm wrong.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1593338,
          "author_name": "mrinath",
          "author_url": "",
          "post_date": "11/23/2021 19:38:09",
          "content": "<p>yes here we are doing that</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1592587": "We all have heard about the F1 score, which is just the harmonic mean of Precision and Recall\n(If you want to revise recall and precision you can see here https://deepai.org/machine-learning-glossary-and-terms/precision-and-recall)\nWhat the F1 score is doing is that it gives equal weightage to Recall and Precision. But sometimes according to our need, we may prefer one of Recall/Precision more than the other.\nA generalised form of F score in terms of 'beta' can be\n`F_beta =((1 + beta^2) * Precision * Recall) / (beta^2 * Precision + Recall)`\nwhere if beta = 1 it gives us the F1 score.\nHere in this task, we are using `beta = 2,`\nhence `F2  score = (5 * Precision * Recall) / (4 * Precision + Recall) `\n F2-measure puts more weight on minimizing false negatives than minimizing false positives. It has the effect of increasing the importance of Recall.\nYou can look more here https://machinelearningmastery.com/fbeta-measure-for-machine-learning/",
    "1593313": "here we are assuming the star fish as the positive and the other species are the negative example.\nPlease correct me if I'm wrong.",
    "1593338": "yes here we are doing that"
  },
  "source": "meta"
}