{
  "id": 71184,
  "title": "F1 score optimization",
  "url": "/competitions/quora-insincere-questions-classification/discussion/71184",
  "author_name": "",
  "post_date": "2018-11-11T05:30:34.215132900Z",
  "votes": 44,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Not sure it has been shared here yet, but @faron had implemented a <a href=\"https://www.kaggle.com/c/instacart-market-basket-analysis/discussion/37221\">very effective FI score optimization</a>  based on the paper <a href=\"https://arxiv.org/abs/1206.4625\">Optimizing F-measure: A Tale of Two Approaches</a>.</p>\n\n<p>I made it run <a href=\"https://www.kaggle.com/cpmpml/f1-score-expectation-maximization-in-o-n/code\">faster using numba</a>.</p>",
  "messages": [
    {
      "id": "419028",
      "postDate": "11/11/2018 05:30:34",
      "content": "<p>Not sure it has been shared here yet, but @faron had implemented a <a href=\"https://www.kaggle.com/c/instacart-market-basket-analysis/discussion/37221\">very effective FI score optimization</a>  based on the paper <a href=\"https://arxiv.org/abs/1206.4625\">Optimizing F-measure: A Tale of Two Approaches</a>.</p>\n\n<p>I made it run <a href=\"https://www.kaggle.com/cpmpml/f1-score-expectation-maximization-in-o-n/code\">faster using numba</a>.</p>",
      "rawMarkdown": "Not sure it has been shared here yet, but @faron had implemented a [very effective FI score optimization][1]  based on the paper [Optimizing F-measure: A Tale of Two Approaches][2].\n\nI made it run [faster using numba][3].\n\n\n  [1]: https://www.kaggle.com/c/instacart-market-basket-analysis/discussion/37221\n  [2]: https://arxiv.org/abs/1206.4625\n  [3]: https://www.kaggle.com/cpmpml/f1-score-expectation-maximization-in-o-n/code",
      "votes": null
    },
    {
      "id": "419255",
      "postDate": "11/11/2018 15:48:52",
      "content": "<p>Could you give an estimation of how much faster ?</p>",
      "rawMarkdown": "Could you give an estimation of how much faster ?",
      "votes": null
    },
    {
      "id": "419289",
      "postDate": "11/11/2018 16:31:43",
      "content": "<p>I don't remember, I did not run it since instacart competition.</p>",
      "rawMarkdown": "I don't remember, I did not run it since instacart competition.",
      "votes": null
    },
    {
      "id": "428992",
      "postDate": "11/28/2018 07:24:04",
      "content": "<p>These two competitions are a little different.</p>\n\n<p>At that competition, it is really hard for us to give threshold of each order.  But we need to get high Mean F1 socre, so Faron's kernel helps a lot.  But for this problem, we only need to tune one threshold, so this may have little help</p>",
      "rawMarkdown": "These two competitions are a little different.\n\nAt that competition, it is really hard for us to give threshold of each order.  But we need to get high Mean F1 socre, so Faron's kernel helps a lot.  But for this problem, we only need to tune one threshold, so this may have little help",
      "votes": null
    },
    {
      "id": "429173",
      "postDate": "11/28/2018 13:18:33",
      "content": "<p>You maybe right, I shared without entering the competition.  I may enter later, we'll see how PLasticc competition goes.</p>",
      "rawMarkdown": "You maybe right, I shared without entering the competition.  I may enter later, we'll see how PLasticc competition goes.",
      "votes": null
    },
    {
      "id": "585843",
      "postDate": "07/28/2019 07:30:51",
      "content": "<p>Any idea how to do F1 score threshold for multiclass classification problem?</p>",
      "rawMarkdown": "Any idea how to do F1 score threshold for multiclass classification problem?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 419255,
      "author_name": "ashutoshmishra1998",
      "author_url": "",
      "post_date": "11/11/2018 15:48:52",
      "content": "<p>Could you give an estimation of how much faster ?</p>",
      "votes": null,
      "replies": [
        {
          "id": 419289,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "11/11/2018 16:31:43",
          "content": "<p>I don't remember, I did not run it since instacart competition.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 428992,
      "author_name": "longyin2",
      "author_url": "",
      "post_date": "11/28/2018 07:24:04",
      "content": "<p>These two competitions are a little different.</p>\n\n<p>At that competition, it is really hard for us to give threshold of each order.  But we need to get high Mean F1 socre, so Faron's kernel helps a lot.  But for this problem, we only need to tune one threshold, so this may have little help</p>",
      "votes": null,
      "replies": [
        {
          "id": 429173,
          "author_name": "cpmpml",
          "author_url": "",
          "post_date": "11/28/2018 13:18:33",
          "content": "<p>You maybe right, I shared without entering the competition.  I may enter later, we'll see how PLasticc competition goes.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 585843,
      "author_name": "shravankoninti",
      "author_url": "",
      "post_date": "07/28/2019 07:30:51",
      "content": "<p>Any idea how to do F1 score threshold for multiclass classification problem?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "419028": "Not sure it has been shared here yet, but @faron had implemented a [very effective FI score optimization][1]  based on the paper [Optimizing F-measure: A Tale of Two Approaches][2].\n\nI made it run [faster using numba][3].\n\n\n  [1]: https://www.kaggle.com/c/instacart-market-basket-analysis/discussion/37221\n  [2]: https://arxiv.org/abs/1206.4625\n  [3]: https://www.kaggle.com/cpmpml/f1-score-expectation-maximization-in-o-n/code",
    "419255": "Could you give an estimation of how much faster ?",
    "419289": "I don't remember, I did not run it since instacart competition.",
    "428992": "These two competitions are a little different.\n\nAt that competition, it is really hard for us to give threshold of each order.  But we need to get high Mean F1 socre, so Faron's kernel helps a lot.  But for this problem, we only need to tune one threshold, so this may have little help",
    "429173": "You maybe right, I shared without entering the competition.  I may enter later, we'll see how PLasticc competition goes.",
    "585843": "Any idea how to do F1 score threshold for multiclass classification problem?"
  },
  "source": "meta"
}