{
  "id": 117077,
  "title": "Zero meter reading values",
  "url": "/competitions/ashrae-energy-prediction/discussion/117077",
  "author_name": "",
  "post_date": "2019-11-13T09:11:20.088097600Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Some meters have strange characteristics. They have a lot of 0 (zero) values. What do you think is the better approach?</p>\n\n<p>a) Do not do anything and train a model leaving them in the dataset.\nb) Filter out 0 values and train a model on the filtered out dataset?\nc) Try to fix the zero values somehow? E.g., replace them with estimated values</p>\n\n<p>Thanks in advanved</p>",
  "messages": [
    {
      "id": "671878",
      "postDate": "11/13/2019 09:11:20",
      "content": "<p>Some meters have strange characteristics. They have a lot of 0 (zero) values. What do you think is the better approach?</p>\n\n<p>a) Do not do anything and train a model leaving them in the dataset.\nb) Filter out 0 values and train a model on the filtered out dataset?\nc) Try to fix the zero values somehow? E.g., replace them with estimated values</p>\n\n<p>Thanks in advanved</p>",
      "rawMarkdown": "Some meters have strange characteristics. They have a lot of 0 (zero) values. What do you think is the better approach?\n\na) Do not do anything and train a model leaving them in the dataset.\nb) Filter out 0 values and train a model on the filtered out dataset?\nc) Try to fix the zero values somehow? E.g., replace them with estimated values\n\nThanks in advanved",
      "votes": null
    },
    {
      "id": "671914",
      "postDate": "11/13/2019 10:10:47",
      "content": "<p>My approach for batches of zeroes is to filter them out - seems to help a bit.</p>",
      "rawMarkdown": "My approach for batches of zeroes is to filter them out - seems to help a bit.",
      "votes": null
    },
    {
      "id": "671926",
      "postDate": "11/13/2019 10:29:33",
      "content": "<p>I haven't tested everything yet but its good to know that there are others thinking about similar approaches.</p>",
      "rawMarkdown": "I haven't tested everything yet but its good to know that there are others thinking about similar approaches.",
      "votes": null
    },
    {
      "id": "675866",
      "postDate": "11/18/2019 17:23:54",
      "content": "<p>I will go with option three.need to replace it with estimated value .For Estimation one can ignore zero values.\nEstimation method can be average of last six months ignoring zero \nOr any other</p>",
      "rawMarkdown": "I will go with option three.need to replace it with estimated value .For Estimation one can ignore zero values.\nEstimation method can be average of last six months ignoring zero \nOr any other",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 671914,
      "author_name": "konradb",
      "author_url": "",
      "post_date": "11/13/2019 10:10:47",
      "content": "<p>My approach for batches of zeroes is to filter them out - seems to help a bit.</p>",
      "votes": null,
      "replies": [
        {
          "id": 671926,
          "author_name": "krinya",
          "author_url": "",
          "post_date": "11/13/2019 10:29:33",
          "content": "<p>I haven't tested everything yet but its good to know that there are others thinking about similar approaches.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 675866,
      "author_name": "smitashah",
      "author_url": "",
      "post_date": "11/18/2019 17:23:54",
      "content": "<p>I will go with option three.need to replace it with estimated value .For Estimation one can ignore zero values.\nEstimation method can be average of last six months ignoring zero \nOr any other</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "671878": "Some meters have strange characteristics. They have a lot of 0 (zero) values. What do you think is the better approach?\n\na) Do not do anything and train a model leaving them in the dataset.\nb) Filter out 0 values and train a model on the filtered out dataset?\nc) Try to fix the zero values somehow? E.g., replace them with estimated values\n\nThanks in advanved",
    "671914": "My approach for batches of zeroes is to filter them out - seems to help a bit.",
    "671926": "I haven't tested everything yet but its good to know that there are others thinking about similar approaches.",
    "675866": "I will go with option three.need to replace it with estimated value .For Estimation one can ignore zero values.\nEstimation method can be average of last six months ignoring zero \nOr any other"
  },
  "source": "meta"
}