{
  "id": 91382,
  "title": "Alignment of acoustic data with time_to_failure",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/91382",
  "author_name": "",
  "post_date": "2019-05-04T01:35:17.813172800Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Maybe this is a very basic question.  When looking at the train data, the very first big peak (which I assume is an earthquake release)  occurs at data point/index 4438688 . Here I find a big spike. The first  lowest value of \"time_to_failure\" however, occurs at index 5656574 (not 0 but very small). The next time to failure is then going up, so a new cycle starts. This would make sense to me, if both events would occur at the same index  (meaning at the same time).  But they don't. What do I miss here?  Thanks for any hint.</p>",
  "messages": [
    {
      "id": "526865",
      "postDate": "05/04/2019 01:35:17",
      "content": "<p>Maybe this is a very basic question.  When looking at the train data, the very first big peak (which I assume is an earthquake release)  occurs at data point/index 4438688 . Here I find a big spike. The first  lowest value of \"time_to_failure\" however, occurs at index 5656574 (not 0 but very small). The next time to failure is then going up, so a new cycle starts. This would make sense to me, if both events would occur at the same index  (meaning at the same time).  But they don't. What do I miss here?  Thanks for any hint.</p>",
      "rawMarkdown": "Maybe this is a very basic question.  When looking at the train data, the very first big peak (which I assume is an earthquake release)  occurs at data point/index 4438688 . Here I find a big spike. The first  lowest value of \"time_to_failure\" however, occurs at index 5656574 (not 0 but very small). The next time to failure is then going up, so a new cycle starts. This would make sense to me, if both events would occur at the same index  (meaning at the same time).  But they don't. What do I miss here?  Thanks for any hint.",
      "votes": null
    },
    {
      "id": "526883",
      "postDate": "05/04/2019 02:28:50",
      "content": "<p>Right, this is one of the issue with this dataset. It has been discussed a number of times in the forum.  The lab quake happens about 0.3 second after the big acoustic spike.</p>",
      "rawMarkdown": "Right, this is one of the issue with this dataset. It has been discussed a number of times in the forum.  The lab quake happens about 0.3 second after the big acoustic spike.",
      "votes": null
    },
    {
      "id": "526918",
      "postDate": "05/04/2019 05:18:14",
      "content": "<p>In one of the reference documents there is discussion about time to start of the quake and time to end - or words to that effect.  </p>\n\n<p>I think the time to end is the data we have - represents when the stress gets to zero.  In a real earthquake this seems that stress gets to zero at the END of what we think is an earthquake.  And in my mind I would want to predict the beginning of the quake, rather than the end.</p>\n\n<p>So the thinking I am using is that the earth quake starts at the big spike and around 0.3 seconds later the stress sensor says its ended.   So our model is to find the \"time to end\".  I am assuming that in their modeling attempts they found the end value easier to work with.</p>\n\n<p>Since I am trying to model in the minor quakes I adjusted the times for the by the 0.3 value that CPMP mentions.  Of course, makes everything confusing late at night when making a prediction.</p>",
      "rawMarkdown": "In one of the reference documents there is discussion about time to start of the quake and time to end - or words to that effect.  \n\nI think the time to end is the data we have - represents when the stress gets to zero.  In a real earthquake this seems that stress gets to zero at the END of what we think is an earthquake.  And in my mind I would want to predict the beginning of the quake, rather than the end.\n\nSo the thinking I am using is that the earth quake starts at the big spike and around 0.3 seconds later the stress sensor says its ended.   So our model is to find the \"time to end\".  I am assuming that in their modeling attempts they found the end value easier to work with.\n\nSince I am trying to model in the minor quakes I adjusted the times for the by the 0.3 value that CPMP mentions.  Of course, makes everything confusing late at night when making a prediction.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 526883,
      "author_name": "cpmpml",
      "author_url": "",
      "post_date": "05/04/2019 02:28:50",
      "content": "<p>Right, this is one of the issue with this dataset. It has been discussed a number of times in the forum.  The lab quake happens about 0.3 second after the big acoustic spike.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 526918,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "05/04/2019 05:18:14",
      "content": "<p>In one of the reference documents there is discussion about time to start of the quake and time to end - or words to that effect.  </p>\n\n<p>I think the time to end is the data we have - represents when the stress gets to zero.  In a real earthquake this seems that stress gets to zero at the END of what we think is an earthquake.  And in my mind I would want to predict the beginning of the quake, rather than the end.</p>\n\n<p>So the thinking I am using is that the earth quake starts at the big spike and around 0.3 seconds later the stress sensor says its ended.   So our model is to find the \"time to end\".  I am assuming that in their modeling attempts they found the end value easier to work with.</p>\n\n<p>Since I am trying to model in the minor quakes I adjusted the times for the by the 0.3 value that CPMP mentions.  Of course, makes everything confusing late at night when making a prediction.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "526865": "Maybe this is a very basic question.  When looking at the train data, the very first big peak (which I assume is an earthquake release)  occurs at data point/index 4438688 . Here I find a big spike. The first  lowest value of \"time_to_failure\" however, occurs at index 5656574 (not 0 but very small). The next time to failure is then going up, so a new cycle starts. This would make sense to me, if both events would occur at the same index  (meaning at the same time).  But they don't. What do I miss here?  Thanks for any hint.",
    "526883": "Right, this is one of the issue with this dataset. It has been discussed a number of times in the forum.  The lab quake happens about 0.3 second after the big acoustic spike.",
    "526918": "In one of the reference documents there is discussion about time to start of the quake and time to end - or words to that effect.  \n\nI think the time to end is the data we have - represents when the stress gets to zero.  In a real earthquake this seems that stress gets to zero at the END of what we think is an earthquake.  And in my mind I would want to predict the beginning of the quake, rather than the end.\n\nSo the thinking I am using is that the earth quake starts at the big spike and around 0.3 seconds later the stress sensor says its ended.   So our model is to find the \"time to end\".  I am assuming that in their modeling attempts they found the end value easier to work with.\n\nSince I am trying to model in the minor quakes I adjusted the times for the by the 0.3 value that CPMP mentions.  Of course, makes everything confusing late at night when making a prediction."
  },
  "source": "meta"
}