{
  "id": 115303,
  "title": "Building Metadata Imputation",
  "url": "/competitions/ashrae-energy-prediction/discussion/115303",
  "author_name": "",
  "post_date": "2019-11-01T18:18:03.029348600Z",
  "votes": -1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>Hi,</p>\n\n<p>Has anyone ever stopped to think about how to impute building metadata values in a way that is not mean and median?</p>\n\n<p><code>sqaure_feet</code>and <code>year_built</code> seem to be very important features and have several NaN values</p>",
  "messages": [
    {
      "id": "663280",
      "postDate": "11/01/2019 18:18:03",
      "content": "<p>Hi,</p>\n\n<p>Has anyone ever stopped to think about how to impute building metadata values in a way that is not mean and median?</p>\n\n<p><code>sqaure_feet</code>and <code>year_built</code> seem to be very important features and have several NaN values</p>",
      "rawMarkdown": "Hi,\n\nHas anyone ever stopped to think about how to impute building metadata values in a way that is not mean and median?\n\n`sqaure_feet`and `year_built` seem to be very important features and have several NaN values",
      "votes": null
    },
    {
      "id": "663299",
      "postDate": "11/01/2019 19:06:19",
      "content": "<p>I think you mean <code>floor_count</code> and <code>year_built</code>. There is no missing<code>square_feet</code> data. If you use tree based algos, you can probably try assigning them -1 and treat them as categorical. There's also model based imputation, mode, or some other constant.</p>",
      "rawMarkdown": "I think you mean `floor_count` and `year_built`. There is no missing` square_feet` data. If you use tree based algos, you can probably try assigning them -1 and treat them as categorical. There's also model based imputation, mode, or some other constant.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 663299,
      "author_name": "teeyee314",
      "author_url": "",
      "post_date": "11/01/2019 19:06:19",
      "content": "<p>I think you mean <code>floor_count</code> and <code>year_built</code>. There is no missing<code>square_feet</code> data. If you use tree based algos, you can probably try assigning them -1 and treat them as categorical. There's also model based imputation, mode, or some other constant.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "663280": "Hi,\n\nHas anyone ever stopped to think about how to impute building metadata values in a way that is not mean and median?\n\n`sqaure_feet`and `year_built` seem to be very important features and have several NaN values",
    "663299": "I think you mean `floor_count` and `year_built`. There is no missing` square_feet` data. If you use tree based algos, you can probably try assigning them -1 and treat them as categorical. There's also model based imputation, mode, or some other constant."
  },
  "source": "meta"
}