{
  "id": 81982,
  "title": "Could someone give an overview of the dataset?",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/81982",
  "author_name": "",
  "post_date": "2019-02-26T16:06:22.306816800Z",
  "votes": 2,
  "comment_count": 3,
  "views": 0,
  "content": "<p>So, I went over the top posts, and seemed to understand the visualization of the dataset, with how the acoustic data seems to <em>spike</em> before an earthquake occurs (with the blue lines <em>peaking</em>), but the actual table of data rather confuses me.</p>\n\n<p>Looking at the data, loaded into my pandas dataframe,\n<img src=\"https://i.imgur.com/DsFrn96.png\" alt=\"enter image description here\"></p>\n\n<p>The <code>acoustic_data</code> needs no explanation, but why doesn't the <code>time_to_failure</code> decrease slightly with each index, as the earthquake draws nearer? I am assuming the dataset is implicitly indexed by time.</p>\n\n<p>Could someone explain it all to me?</p>",
  "messages": [
    {
      "id": "478773",
      "postDate": "02/26/2019 16:06:22",
      "content": "<p>So, I went over the top posts, and seemed to understand the visualization of the dataset, with how the acoustic data seems to <em>spike</em> before an earthquake occurs (with the blue lines <em>peaking</em>), but the actual table of data rather confuses me.</p>\n\n<p>Looking at the data, loaded into my pandas dataframe,\n<img src=\"https://i.imgur.com/DsFrn96.png\" alt=\"enter image description here\"></p>\n\n<p>The <code>acoustic_data</code> needs no explanation, but why doesn't the <code>time_to_failure</code> decrease slightly with each index, as the earthquake draws nearer? I am assuming the dataset is implicitly indexed by time.</p>\n\n<p>Could someone explain it all to me?</p>",
      "rawMarkdown": "So, I went over the top posts, and seemed to understand the visualization of the dataset, with how the acoustic data seems to *spike* before an earthquake occurs (with the blue lines *peaking*), but the actual table of data rather confuses me.\n\nLooking at the data, loaded into my pandas dataframe,\n![enter image description here][1]\n\nThe `acoustic_data` needs no explanation, but why doesn't the `time_to_failure` decrease slightly with each index, as the earthquake draws nearer? I am assuming the dataset is implicitly indexed by time.\n\nCould someone explain it all to me?\n\n  [1]: https://i.imgur.com/DsFrn96.png",
      "votes": null
    },
    {
      "id": "478786",
      "postDate": "02/26/2019 16:33:21",
      "content": "<p>Change how many digits to display after the decimal point. Also, 'additional info' topic  has some very interesting discussions- read it's all posts, not only the top rated. \nWhat blue lines?</p>",
      "rawMarkdown": "Change how many digits to display after the decimal point. Also, 'additional info' topic  has some very interesting discussions- read it's all posts, not only the top rated. \nWhat blue lines?",
      "votes": null
    },
    {
      "id": "478853",
      "postDate": "02/26/2019 17:57:50",
      "content": "<p>Forgive me, that was quite amateurish, as the last graph I saw was not the one in the top post, but in the youtube vid someone else posted in this forum, with the graph here <a href=\"https://i.imgur.com/D86VI2n.png\">https://i.imgur.com/D86VI2n.png</a></p>\n\n<p>And on the precision point, thanks for pointing that out &gt;.&lt; , now I see the countdown, so to speak.</p>",
      "rawMarkdown": "Forgive me, that was quite amateurish, as the last graph I saw was not the one in the top post, but in the youtube vid someone else posted in this forum, with the graph here https://i.imgur.com/D86VI2n.png\n\nAnd on the precision point, thanks for pointing that out &gt;.&lt; , now I see the countdown, so to speak.",
      "votes": null
    },
    {
      "id": "478895",
      "postDate": "02/26/2019 19:02:05",
      "content": "<p>Yes, those spikes in the training data are , (poetically speaking) unrealised earthquakes. There the system doesn't failure (doesn't reach critical state) and recover (the experiment continue). I noticed that in each earthquake cycle that the last time to failure value for the last bin(see 'aditional info topic's) is always less than the length of those bins (about a millisecond). That makes me to conclude that if an earthquake /system failure happens within a bin, that particular bin would be incomplete and is not included in the available data. So, we never see actual earthquake.</p>",
      "rawMarkdown": "Yes, those spikes in the training data are , (poetically speaking) unrealised earthquakes. There the system doesn't failure (doesn't reach critical state) and recover (the experiment continue). I noticed that in each earthquake cycle that the last time to failure value for the last bin(see 'aditional info topic's) is always less than the length of those bins (about a millisecond). That makes me to conclude that if an earthquake /system failure happens within a bin, that particular bin would be incomplete and is not included in the available data. So, we never see actual earthquake.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 478786,
      "author_name": "petya5q",
      "author_url": "",
      "post_date": "02/26/2019 16:33:21",
      "content": "<p>Change how many digits to display after the decimal point. Also, 'additional info' topic  has some very interesting discussions- read it's all posts, not only the top rated. \nWhat blue lines?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 478853,
      "author_name": "mike333",
      "author_url": "",
      "post_date": "02/26/2019 17:57:50",
      "content": "<p>Forgive me, that was quite amateurish, as the last graph I saw was not the one in the top post, but in the youtube vid someone else posted in this forum, with the graph here <a href=\"https://i.imgur.com/D86VI2n.png\">https://i.imgur.com/D86VI2n.png</a></p>\n\n<p>And on the precision point, thanks for pointing that out &gt;.&lt; , now I see the countdown, so to speak.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 478895,
      "author_name": "petya5q",
      "author_url": "",
      "post_date": "02/26/2019 19:02:05",
      "content": "<p>Yes, those spikes in the training data are , (poetically speaking) unrealised earthquakes. There the system doesn't failure (doesn't reach critical state) and recover (the experiment continue). I noticed that in each earthquake cycle that the last time to failure value for the last bin(see 'aditional info topic's) is always less than the length of those bins (about a millisecond). That makes me to conclude that if an earthquake /system failure happens within a bin, that particular bin would be incomplete and is not included in the available data. So, we never see actual earthquake.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "478773": "So, I went over the top posts, and seemed to understand the visualization of the dataset, with how the acoustic data seems to *spike* before an earthquake occurs (with the blue lines *peaking*), but the actual table of data rather confuses me.\n\nLooking at the data, loaded into my pandas dataframe,\n![enter image description here][1]\n\nThe `acoustic_data` needs no explanation, but why doesn't the `time_to_failure` decrease slightly with each index, as the earthquake draws nearer? I am assuming the dataset is implicitly indexed by time.\n\nCould someone explain it all to me?\n\n  [1]: https://i.imgur.com/DsFrn96.png",
    "478786": "Change how many digits to display after the decimal point. Also, 'additional info' topic  has some very interesting discussions- read it's all posts, not only the top rated. \nWhat blue lines?",
    "478853": "Forgive me, that was quite amateurish, as the last graph I saw was not the one in the top post, but in the youtube vid someone else posted in this forum, with the graph here https://i.imgur.com/D86VI2n.png\n\nAnd on the precision point, thanks for pointing that out &gt;.&lt; , now I see the countdown, so to speak.",
    "478895": "Yes, those spikes in the training data are , (poetically speaking) unrealised earthquakes. There the system doesn't failure (doesn't reach critical state) and recover (the experiment continue). I noticed that in each earthquake cycle that the last time to failure value for the last bin(see 'aditional info topic's) is always less than the length of those bins (about a millisecond). That makes me to conclude that if an earthquake /system failure happens within a bin, that particular bin would be incomplete and is not included in the available data. So, we never see actual earthquake."
  },
  "source": "meta"
}