{
  "id": 124582,
  "title": "A question about Outliers",
  "url": "/competitions/ashrae-energy-prediction/discussion/124582",
  "author_name": "",
  "post_date": "2020-01-05T04:07:05.134743300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I really enjoyed this competition and I have a question about outliers.\nis there any blog or kernel about outliers? I mean, a good tutorial you recommend, there are plenty of blogs on the internet but I don't know which is effective.\nNow I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero) or by plotting all values and identifying non-normal values.\nAfter knowing the outliers index, do I just have to delete them? or there are other options to proceed.</p>\n\n<p>Thanks so much for your help :D</p>",
  "messages": [
    {
      "id": "710671",
      "postDate": "01/05/2020 04:07:05",
      "content": "<p>I really enjoyed this competition and I have a question about outliers.\nis there any blog or kernel about outliers? I mean, a good tutorial you recommend, there are plenty of blogs on the internet but I don't know which is effective.\nNow I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero) or by plotting all values and identifying non-normal values.\nAfter knowing the outliers index, do I just have to delete them? or there are other options to proceed.</p>\n\n<p>Thanks so much for your help :D</p>",
      "rawMarkdown": "I really enjoyed this competition and I have a question about outliers.\nis there any blog or kernel about outliers? I mean, a good tutorial you recommend, there are plenty of blogs on the internet but I don't know which is effective.\nNow I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero) or by plotting all values and identifying non-normal values.\nAfter knowing the outliers index, do I just have to delete them? or there are other options to proceed.\n\nThanks so much for your help :D",
      "votes": null
    },
    {
      "id": "710750",
      "postDate": "01/05/2020 07:03:44",
      "content": "<p>This post shows how I dealt with outliers.\n<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471</a></p>\n\n<blockquote>\n  <p>Now I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero)</p>\n</blockquote>\n\n<p>I think EDA was more important than logic in this competition because those 0 electricity meter readings probably exist in test too.</p>\n\n<blockquote>\n  <p>After knowing the outliers index, do I just have to delete them? or there are other options to proceed.</p>\n</blockquote>\n\n<p>We tried giving outliers =~0.0001 sample weight for training, but it was better to drop them.</p>",
      "rawMarkdown": "This post shows how I dealt with outliers.\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471\n\n&gt; Now I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero)\n\nI think EDA was more important than logic in this competition because those 0 electricity meter readings probably exist in test too.\n\n&gt; After knowing the outliers index, do I just have to delete them? or there are other options to proceed.\n\nWe tried giving outliers =~0.0001 sample weight for training, but it was better to drop them.",
      "votes": null
    },
    {
      "id": "711151",
      "postDate": "01/05/2020 18:43:20",
      "content": "<p>Thanks so much for your answer :D</p>",
      "rawMarkdown": "Thanks so much for your answer :D",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 710750,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "01/05/2020 07:03:44",
      "content": "<p>This post shows how I dealt with outliers.\n<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471</a></p>\n\n<blockquote>\n  <p>Now I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero)</p>\n</blockquote>\n\n<p>I think EDA was more important than logic in this competition because those 0 electricity meter readings probably exist in test too.</p>\n\n<blockquote>\n  <p>After knowing the outliers index, do I just have to delete them? or there are other options to proceed.</p>\n</blockquote>\n\n<p>We tried giving outliers =~0.0001 sample weight for training, but it was better to drop them.</p>",
      "votes": null,
      "replies": [
        {
          "id": 711151,
          "author_name": "johnt666",
          "author_url": "",
          "post_date": "01/05/2020 18:43:20",
          "content": "<p>Thanks so much for your answer :D</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "710671": "I really enjoyed this competition and I have a question about outliers.\nis there any blog or kernel about outliers? I mean, a good tutorial you recommend, there are plenty of blogs on the internet but I don't know which is effective.\nNow I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero) or by plotting all values and identifying non-normal values.\nAfter knowing the outliers index, do I just have to delete them? or there are other options to proceed.\n\nThanks so much for your help :D",
    "710750": "This post shows how I dealt with outliers.\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/122471\n\n&gt; Now I think I just have to identify values that are not logical by concept (for instance, that electrical meter reading can't be zero)\n\nI think EDA was more important than logic in this competition because those 0 electricity meter readings probably exist in test too.\n\n&gt; After knowing the outliers index, do I just have to delete them? or there are other options to proceed.\n\nWe tried giving outliers =~0.0001 sample weight for training, but it was better to drop them.",
    "711151": "Thanks so much for your answer :D"
  },
  "source": "meta"
}