{
  "id": 119670,
  "title": "Should there be NaNs in the meter_reading submission?",
  "url": "/competitions/ashrae-energy-prediction/discussion/119670",
  "author_name": "",
  "post_date": "2019-11-30T11:40:10.430649200Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Hi all,</p>\n\n<p>I'm late to the competition and am wondering if there should be NaN in the meter reading submission because the corresponding weather data is unavailable? or if I did something wrong?</p>\n\n<p>Thanks!</p>",
  "messages": [
    {
      "id": "684735",
      "postDate": "11/30/2019 11:40:10",
      "content": "<p>Hi all,</p>\n\n<p>I'm late to the competition and am wondering if there should be NaN in the meter reading submission because the corresponding weather data is unavailable? or if I did something wrong?</p>\n\n<p>Thanks!</p>",
      "rawMarkdown": "Hi all,\n\nI'm late to the competition and am wondering if there should be NaN in the meter reading submission because the corresponding weather data is unavailable? or if I did something wrong?\n\nThanks!",
      "votes": null
    },
    {
      "id": "685043",
      "postDate": "11/30/2019 21:27:36",
      "content": "<p>No, even if your model predicts NaN, you should introduce a numeric, e.g. 0, otherwise the submission system will give you an error. If your model needs weather data that is unavailable try using a flag for Nan (e.g. -999999), or a mean/ median value or interpolate the missing values from the ones that are available. </p>",
      "rawMarkdown": "No, even if your model predicts NaN, you should introduce a numeric, e.g. 0, otherwise the submission system will give you an error. If your model needs weather data that is unavailable try using a flag for Nan (e.g. -999999), or a mean/ median value or interpolate the missing values from the ones that are available.",
      "votes": null
    },
    {
      "id": "685154",
      "postDate": "12/01/2019 03:34:26",
      "content": "<p>Interpolation isn't too hard. Check out <a href=\"https://www.kaggle.com/purist1024/strategy-evaluation-what-helps-and-by-how-much\">this kernel</a>. It gives a practical implementation of interpolation but also shows that (surprisingly) it only helps by about half a percent.</p>",
      "rawMarkdown": "Interpolation isn't too hard. Check out [this kernel](https://www.kaggle.com/purist1024/strategy-evaluation-what-helps-and-by-how-much). It gives a practical implementation of interpolation but also shows that (surprisingly) it only helps by about half a percent.",
      "votes": null
    },
    {
      "id": "685269",
      "postDate": "12/01/2019 10:03:01",
      "content": "<p>Thank you both!</p>",
      "rawMarkdown": "Thank you both!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 685043,
      "author_name": "panosc",
      "author_url": "",
      "post_date": "11/30/2019 21:27:36",
      "content": "<p>No, even if your model predicts NaN, you should introduce a numeric, e.g. 0, otherwise the submission system will give you an error. If your model needs weather data that is unavailable try using a flag for Nan (e.g. -999999), or a mean/ median value or interpolate the missing values from the ones that are available. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 685154,
      "author_name": "purist1024",
      "author_url": "",
      "post_date": "12/01/2019 03:34:26",
      "content": "<p>Interpolation isn't too hard. Check out <a href=\"https://www.kaggle.com/purist1024/strategy-evaluation-what-helps-and-by-how-much\">this kernel</a>. It gives a practical implementation of interpolation but also shows that (surprisingly) it only helps by about half a percent.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 685269,
      "author_name": "lunatics",
      "author_url": "",
      "post_date": "12/01/2019 10:03:01",
      "content": "<p>Thank you both!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "684735": "Hi all,\n\nI'm late to the competition and am wondering if there should be NaN in the meter reading submission because the corresponding weather data is unavailable? or if I did something wrong?\n\nThanks!",
    "685043": "No, even if your model predicts NaN, you should introduce a numeric, e.g. 0, otherwise the submission system will give you an error. If your model needs weather data that is unavailable try using a flag for Nan (e.g. -999999), or a mean/ median value or interpolate the missing values from the ones that are available.",
    "685154": "Interpolation isn't too hard. Check out [this kernel](https://www.kaggle.com/purist1024/strategy-evaluation-what-helps-and-by-how-much). It gives a practical implementation of interpolation but also shows that (surprisingly) it only helps by about half a percent.",
    "685269": "Thank you both!"
  },
  "source": "meta"
}