{
  "id": 92578,
  "title": "Why do we slit at 16 points eathquake?",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/92578",
  "author_name": "",
  "post_date": "2019-05-18T04:29:28.268984200Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Kernel : <a href=\"https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests\">https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests</a>\nWhy do we slit at 16 points eathquake?\nCan I split at point other?\nPlease explain for me</p>",
  "messages": [
    {
      "id": "532935",
      "postDate": "05/18/2019 04:29:28",
      "content": "<p>Kernel : <a href=\"https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests\">https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests</a>\nWhy do we slit at 16 points eathquake?\nCan I split at point other?\nPlease explain for me</p>",
      "rawMarkdown": "Kernel : https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests\nWhy do we slit at 16 points eathquake?\nCan I split at point other?\nPlease explain for me",
      "votes": null
    },
    {
      "id": "532956",
      "postDate": "05/18/2019 05:33:37",
      "content": "<p>The assumption or idea is that there are 16 <em>independent</em> earthquakes in train. Independence here means that one earthquake does not affect another one. When the experiment resets and <code>time_to_failure</code> shoots back up from near 0, then you basically have a new continuous series of <code>acoustic_data</code> with accompanying target variable, <code>time_to_failure</code>. Splitting up the continuous series of data makes sense in this context. The next step is figuring out what a good validation approach is in light of having 16 separate earthquakes.</p>",
      "rawMarkdown": "The assumption or idea is that there are 16 _independent_ earthquakes in train. Independence here means that one earthquake does not affect another one. When the experiment resets and `time_to_failure` shoots back up from near 0, then you basically have a new continuous series of `acoustic_data` with accompanying target variable, `time_to_failure`. Splitting up the continuous series of data makes sense in this context. The next step is figuring out what a good validation approach is in light of having 16 separate earthquakes.",
      "votes": null
    },
    {
      "id": "532999",
      "postDate": "05/18/2019 08:05:49",
      "content": "<blockquote>\n  <p>Why do we slit at 16 points eathquake?</p>\n</blockquote>\n\n<p>Because someone did and everybody followed.</p>\n\n<blockquote>\n  <p>Can I split at point other?</p>\n</blockquote>\n\n<p>Of course you can.  Some people use unshuffled K fold, or shuffled K fold.</p>",
      "rawMarkdown": "&gt; Why do we slit at 16 points eathquake?\n\nBecause someone did and everybody followed.\n\n&gt;  Can I split at point other?\n\nOf course you can.  Some people use unshuffled K fold, or shuffled K fold.",
      "votes": null
    },
    {
      "id": "533034",
      "postDate": "05/18/2019 10:00:40",
      "content": "<p>Thank you explained.\nCan I split at points other (ex points random) on the train data? </p>",
      "rawMarkdown": "Thank you explained.\nCan I split at points other (ex points random) on the train data?",
      "votes": null
    },
    {
      "id": "533035",
      "postDate": "05/18/2019 10:01:23",
      "content": "<p>Thank you very much</p>",
      "rawMarkdown": "Thank you very much",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 532956,
      "author_name": "teeyee314",
      "author_url": "",
      "post_date": "05/18/2019 05:33:37",
      "content": "<p>The assumption or idea is that there are 16 <em>independent</em> earthquakes in train. Independence here means that one earthquake does not affect another one. When the experiment resets and <code>time_to_failure</code> shoots back up from near 0, then you basically have a new continuous series of <code>acoustic_data</code> with accompanying target variable, <code>time_to_failure</code>. Splitting up the continuous series of data makes sense in this context. The next step is figuring out what a good validation approach is in light of having 16 separate earthquakes.</p>",
      "votes": null,
      "replies": [
        {
          "id": 533034,
          "author_name": "dat12345",
          "author_url": "",
          "post_date": "05/18/2019 10:00:40",
          "content": "<p>Thank you explained.\nCan I split at points other (ex points random) on the train data? </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 532999,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "05/18/2019 08:05:49",
      "content": "<blockquote>\n  <p>Why do we slit at 16 points eathquake?</p>\n</blockquote>\n\n<p>Because someone did and everybody followed.</p>\n\n<blockquote>\n  <p>Can I split at point other?</p>\n</blockquote>\n\n<p>Of course you can.  Some people use unshuffled K fold, or shuffled K fold.</p>",
      "votes": null,
      "replies": [
        {
          "id": 533035,
          "author_name": "dat12345",
          "author_url": "",
          "post_date": "05/18/2019 10:01:23",
          "content": "<p>Thank you very much</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "532935": "Kernel : https://www.kaggle.com/mayer79/earthquake-forecast-with-random-forests\nWhy do we slit at 16 points eathquake?\nCan I split at point other?\nPlease explain for me",
    "532956": "The assumption or idea is that there are 16 _independent_ earthquakes in train. Independence here means that one earthquake does not affect another one. When the experiment resets and `time_to_failure` shoots back up from near 0, then you basically have a new continuous series of `acoustic_data` with accompanying target variable, `time_to_failure`. Splitting up the continuous series of data makes sense in this context. The next step is figuring out what a good validation approach is in light of having 16 separate earthquakes.",
    "532999": "&gt; Why do we slit at 16 points eathquake?\n\nBecause someone did and everybody followed.\n\n&gt;  Can I split at point other?\n\nOf course you can.  Some people use unshuffled K fold, or shuffled K fold.",
    "533034": "Thank you explained.\nCan I split at points other (ex points random) on the train data?",
    "533035": "Thank you very much"
  },
  "source": "meta"
}