{
  "id": 114538,
  "title": "There are no gaps in test set",
  "url": "/competitions/ashrae-energy-prediction/discussion/114538",
  "author_name": "",
  "post_date": "2019-10-27T06:36:44.783234600Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>It is expected that there are gaps (lack of data in certain timestamps for certain buildings and meters) in train and test data.</p>\n\n<p>There may be many reasons why there are gaps: </p>\n\n<ul>\n<li>The meter is off.</li>\n<li>Communication error with meter.</li>\n<li>The software that collects data (or part of it) is hung.</li>\n<li>The server is off.</li>\n</ul>\n\n<p>I did a <a href=\"https://www.kaggle.com/juanmah/ashrae-gaps\">notebook</a> analysing the gaps in the train and test set. Where you can see the results:</p>\n\n<ul>\n<li>In train set, there are 689,820 gaps for a total of 20,216,100 samples. The ratio of gaps is: 3.41%.</li>\n<li><strong>In test set, there is no gap at all.</strong></li>\n</ul>\n\n<p>It is very unlikely that there is no gap in the test set. I expected that there would be gaps in the test set, in which no prediction would have to be made.</p>\n\n<p>I ask myself a few questions for which I have no answers:</p>\n\n<ul>\n<li>Why there are no gaps?</li>\n<li>What values will there be instead of the gaps?</li>\n<li>How can it affect the score?</li>\n</ul>",
  "messages": [
    {
      "id": "659140",
      "postDate": "10/27/2019 06:36:44",
      "content": "<p>It is expected that there are gaps (lack of data in certain timestamps for certain buildings and meters) in train and test data.</p>\n\n<p>There may be many reasons why there are gaps: </p>\n\n<ul>\n<li>The meter is off.</li>\n<li>Communication error with meter.</li>\n<li>The software that collects data (or part of it) is hung.</li>\n<li>The server is off.</li>\n</ul>\n\n<p>I did a <a href=\"https://www.kaggle.com/juanmah/ashrae-gaps\">notebook</a> analysing the gaps in the train and test set. Where you can see the results:</p>\n\n<ul>\n<li>In train set, there are 689,820 gaps for a total of 20,216,100 samples. The ratio of gaps is: 3.41%.</li>\n<li><strong>In test set, there is no gap at all.</strong></li>\n</ul>\n\n<p>It is very unlikely that there is no gap in the test set. I expected that there would be gaps in the test set, in which no prediction would have to be made.</p>\n\n<p>I ask myself a few questions for which I have no answers:</p>\n\n<ul>\n<li>Why there are no gaps?</li>\n<li>What values will there be instead of the gaps?</li>\n<li>How can it affect the score?</li>\n</ul>",
      "rawMarkdown": "It is expected that there are gaps (lack of data in certain timestamps for certain buildings and meters) in train and test data.\n\nThere may be many reasons why there are gaps: \n\n- The meter is off.\n- Communication error with meter.\n- The software that collects data (or part of it) is hung.\n- The server is off.\n\nI did a [notebook](https://www.kaggle.com/juanmah/ashrae-gaps) analysing the gaps in the train and test set. Where you can see the results:\n\n- In train set, there are 689,820 gaps for a total of 20,216,100 samples. The ratio of gaps is: 3.41%.\n- **In test set, there is no gap at all.**\n\nIt is very unlikely that there is no gap in the test set. I expected that there would be gaps in the test set, in which no prediction would have to be made.\n\nI ask myself a few questions for which I have no answers:\n\n- Why there are no gaps?\n- What values will there be instead of the gaps?\n- How can it affect the score?",
      "votes": null
    },
    {
      "id": "659143",
      "postDate": "10/27/2019 06:44:33",
      "content": "<p>There was a comment by the sponsor that if any gaps existed in test they would not be scored.  I think that meant that gaps existed but the test set was built for the full hours in 2018 for every building.  </p>",
      "rawMarkdown": "There was a comment by the sponsor that if any gaps existed in test they would not be scored.  I think that meant that gaps existed but the test set was built for the full hours in 2018 for every building.",
      "votes": null
    },
    {
      "id": "659181",
      "postDate": "10/27/2019 08:14:02",
      "content": "<p>That makes sense.</p>\n\n<p>I found it in data description:</p>\n\n<blockquote>\n  <p>Gaps in the test set are not revealed or scored.</p>\n</blockquote>\n\n<p>Thanks.</p>",
      "rawMarkdown": "That makes sense.\n\nI found it in data description:\n\n&gt; Gaps in the test set are not revealed or scored.\n\nThanks.",
      "votes": null
    },
    {
      "id": "664980",
      "postDate": "11/04/2019 13:44:36",
      "content": "<p>Some predictions will not be scored.</p>",
      "rawMarkdown": "Some predictions will not be scored.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 659143,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "10/27/2019 06:44:33",
      "content": "<p>There was a comment by the sponsor that if any gaps existed in test they would not be scored.  I think that meant that gaps existed but the test set was built for the full hours in 2018 for every building.  </p>",
      "votes": null,
      "replies": [
        {
          "id": 659181,
          "author_name": "juanmah",
          "author_url": "",
          "post_date": "10/27/2019 08:14:02",
          "content": "<p>That makes sense.</p>\n\n<p>I found it in data description:</p>\n\n<blockquote>\n  <p>Gaps in the test set are not revealed or scored.</p>\n</blockquote>\n\n<p>Thanks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 664980,
      "author_name": "longyin2",
      "author_url": "",
      "post_date": "11/04/2019 13:44:36",
      "content": "<p>Some predictions will not be scored.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "659140": "It is expected that there are gaps (lack of data in certain timestamps for certain buildings and meters) in train and test data.\n\nThere may be many reasons why there are gaps: \n\n- The meter is off.\n- Communication error with meter.\n- The software that collects data (or part of it) is hung.\n- The server is off.\n\nI did a [notebook](https://www.kaggle.com/juanmah/ashrae-gaps) analysing the gaps in the train and test set. Where you can see the results:\n\n- In train set, there are 689,820 gaps for a total of 20,216,100 samples. The ratio of gaps is: 3.41%.\n- **In test set, there is no gap at all.**\n\nIt is very unlikely that there is no gap in the test set. I expected that there would be gaps in the test set, in which no prediction would have to be made.\n\nI ask myself a few questions for which I have no answers:\n\n- Why there are no gaps?\n- What values will there be instead of the gaps?\n- How can it affect the score?",
    "659143": "There was a comment by the sponsor that if any gaps existed in test they would not be scored.  I think that meant that gaps existed but the test set was built for the full hours in 2018 for every building.",
    "659181": "That makes sense.\n\nI found it in data description:\n\n&gt; Gaps in the test set are not revealed or scored.\n\nThanks.",
    "664980": "Some predictions will not be scored."
  },
  "source": "meta"
}