{
  "id": 73024,
  "title": "How to handle these too many outliers in target values?",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/73024",
  "author_name": "",
  "post_date": "2018-11-29T04:14:10.563714400Z",
  "votes": 3,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hey, mates:</p>\n\n<p>I have do some work for handle outliers in target values as following link, \n<a href=\"https://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173\">https://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173</a></p>\n\n<p>Also, we can discuss more methods or kagglers' tricks for this at here.</p>\n\n<p>Thanks.</p>",
  "messages": [
    {
      "id": "429611",
      "postDate": "11/29/2018 04:14:10",
      "content": "<p>Hey, mates:</p>\n\n<p>I have do some work for handle outliers in target values as following link, \n<a href=\"https://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173\">https://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173</a></p>\n\n<p>Also, we can discuss more methods or kagglers' tricks for this at here.</p>\n\n<p>Thanks.</p>",
      "rawMarkdown": "Hey, mates:\n\nI have do some work for handle outliers in target values as following link, \nhttps://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173\n\nAlso, we can discuss more methods or kagglers' tricks for this at here.\n\nThanks.",
      "votes": null
    },
    {
      "id": "429805",
      "postDate": "11/29/2018 11:18:18",
      "content": "<p>Hello Duo,</p>\n\n<p>I've also seen those outliers.. I am bit confused with those values.. \nFor now I used some normalization on those values but i'm not sure it is the best way to handle this.</p>",
      "rawMarkdown": "Hello Duo,\n\nI've also seen those outliers.. I am bit confused with those values.. \nFor now I used some normalization on those values but i'm not sure it is the best way to handle this.",
      "votes": null
    },
    {
      "id": "429853",
      "postDate": "11/29/2018 13:13:25",
      "content": "<p>Yes, I think it may be a big challenge for this competition.</p>",
      "rawMarkdown": "Yes, I think it may be a big challenge for this competition.",
      "votes": null
    },
    {
      "id": "429963",
      "postDate": "11/29/2018 15:51:58",
      "content": "<p>One of the ways is simply clipping too high values. Like values at 95-99 percentile or values higher than mean + 3 sigma. The same for too low values.</p>",
      "rawMarkdown": "One of the ways is simply clipping too high values. Like values at 95-99 percentile or values higher than mean + 3 sigma. The same for too low values.",
      "votes": null
    },
    {
      "id": "430000",
      "postDate": "11/29/2018 16:33:36",
      "content": "<p>You may remove these values (not the entire row), and then impute them again.</p>",
      "rawMarkdown": "You may remove these values (not the entire row), and then impute them again.",
      "votes": null
    },
    {
      "id": "459420",
      "postDate": "01/21/2019 17:33:38",
      "content": "<p>By removing those values, do you mean replacing those values with a clip-off value?</p>",
      "rawMarkdown": "By removing those values, do you mean replacing those values with a clip-off value?",
      "votes": null
    },
    {
      "id": "459523",
      "postDate": "01/21/2019 23:06:13",
      "content": "<p>Exactly.</p>",
      "rawMarkdown": "Exactly.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 429805,
      "author_name": "xtelima",
      "author_url": "",
      "post_date": "11/29/2018 11:18:18",
      "content": "<p>Hello Duo,</p>\n\n<p>I've also seen those outliers.. I am bit confused with those values.. \nFor now I used some normalization on those values but i'm not sure it is the best way to handle this.</p>",
      "votes": null,
      "replies": [
        {
          "id": 429853,
          "author_name": "classtag",
          "author_url": "",
          "post_date": "11/29/2018 13:13:25",
          "content": "<p>Yes, I think it may be a big challenge for this competition.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 429963,
      "author_name": "artgor",
      "author_url": "",
      "post_date": "11/29/2018 15:51:58",
      "content": "<p>One of the ways is simply clipping too high values. Like values at 95-99 percentile or values higher than mean + 3 sigma. The same for too low values.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 430000,
      "author_name": "roydatascience",
      "author_url": "",
      "post_date": "11/29/2018 16:33:36",
      "content": "<p>You may remove these values (not the entire row), and then impute them again.</p>",
      "votes": null,
      "replies": [
        {
          "id": 459420,
          "author_name": "pukkinming",
          "author_url": "",
          "post_date": "01/21/2019 17:33:38",
          "content": "<p>By removing those values, do you mean replacing those values with a clip-off value?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 459523,
          "author_name": "roydatascience",
          "author_url": "",
          "post_date": "01/21/2019 23:06:13",
          "content": "<p>Exactly.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "429611": "Hey, mates:\n\nI have do some work for handle outliers in target values as following link, \nhttps://www.kaggle.com/classtag/how-to-handle-this-most-outliers-in-target-values?scriptVersionId=7878173\n\nAlso, we can discuss more methods or kagglers' tricks for this at here.\n\nThanks.",
    "429805": "Hello Duo,\n\nI've also seen those outliers.. I am bit confused with those values.. \nFor now I used some normalization on those values but i'm not sure it is the best way to handle this.",
    "429853": "Yes, I think it may be a big challenge for this competition.",
    "429963": "One of the ways is simply clipping too high values. Like values at 95-99 percentile or values higher than mean + 3 sigma. The same for too low values.",
    "430000": "You may remove these values (not the entire row), and then impute them again.",
    "459420": "By removing those values, do you mean replacing those values with a clip-off value?",
    "459523": "Exactly."
  },
  "source": "meta"
}