{
  "id": 190682,
  "title": "What is time to eruption measured in and how many time steps between measurements in the signal data?",
  "url": "/competitions/predict-volcanic-eruptions-ingv-oe/discussion/190682",
  "author_name": "",
  "post_date": "2020-10-13T01:14:52.784974300Z",
  "votes": 19,
  "comment_count": 7,
  "views": 0,
  "content": "<p>Is time to eruption minutes, seconds, milliseconds, microseconds? How many time-steps between the measurements in the signal data?</p>\n<p>I cannot seem to find that metric anywhere in the documentation, if anybody has found it please inform!</p>\n<p>Thanks!!</p>\n<p><strong>I see that it says that the observations are 10 min long, there are 60,000 rows per segment. 600 seconds per second. Meaning that the observations must be every 1/100 of a second. Let me know if I did that wrong.</strong></p>\n<p>Can we assume that the time to eruption is also in 1/100 of seconds (i.e centiseconds)</p>",
  "messages": [
    {
      "id": "1047815",
      "postDate": "10/13/2020 01:14:52",
      "content": "<p>Is time to eruption minutes, seconds, milliseconds, microseconds? How many time-steps between the measurements in the signal data?</p>\n<p>I cannot seem to find that metric anywhere in the documentation, if anybody has found it please inform!</p>\n<p>Thanks!!</p>\n<p><strong>I see that it says that the observations are 10 min long, there are 60,000 rows per segment. 600 seconds per second. Meaning that the observations must be every 1/100 of a second. Let me know if I did that wrong.</strong></p>\n<p>Can we assume that the time to eruption is also in 1/100 of seconds (i.e centiseconds)</p>",
      "rawMarkdown": "Is time to eruption minutes, seconds, milliseconds, microseconds? How many time-steps between the measurements in the signal data?\n\nI cannot seem to find that metric anywhere in the documentation, if anybody has found it please inform!\n\nThanks!!\n\n**I see that it says that the observations are 10 min long, there are 60,000 rows per segment. 600 seconds per second. Meaning that the observations must be every 1/100 of a second. Let me know if I did that wrong.**\n\nCan we assume that the time to eruption is also in 1/100 of seconds (i.e centiseconds)",
      "votes": null
    },
    {
      "id": "1047941",
      "postDate": "10/13/2020 04:34:30",
      "content": "<p>At first look it seemed to me like a regression problem, but reading your thread made me realise that it more similar to a time sequence prediction problem. Only that instead of sequence we are to predict the time and expected sequence corresponding to volcanic eruption. Thanks.</p>",
      "rawMarkdown": "At first look it seemed to me like a regression problem, but reading your thread made me realise that it more similar to a time sequence prediction problem. Only that instead of sequence we are to predict the time and expected sequence corresponding to volcanic eruption. Thanks.",
      "votes": null
    },
    {
      "id": "1048200",
      "postDate": "10/13/2020 09:22:10",
      "content": "<p>Yes, you can assume the sampling rate at 100Hz and the time_to_eruption as the number of samples before the next eruption.</p>",
      "rawMarkdown": "Yes, you can assume the sampling rate at 100Hz and the time_to_eruption as the number of samples before the next eruption.",
      "votes": null
    },
    {
      "id": "1066288",
      "postDate": "11/01/2020 15:30:31",
      "content": "<p>There are definitely overlapping segments right? Am I doing something wrong? If there are 4,000 segments each 60,000 centiseconds that means the <code>time_to_eruption</code> would have to span 240,000,000 centiseconds for no overlapping segments. Our competition only spans 40,000,000 centi seconds or 112 hours.</p>\n<p>And there are overlapping segments where sensors are on/off which means there are duplicates of each sensor? Are there multiple measurement points or multiple eruptions?</p>",
      "rawMarkdown": "There are definitely overlapping segments right? Am I doing something wrong? If there are 4,000 segments each 60,000 centiseconds that means the `time_to_eruption` would have to span 240,000,000 centiseconds for no overlapping segments. Our competition only spans 40,000,000 centi seconds or 112 hours.\n\n\nAnd there are overlapping segments where sensors are on/off which means there are duplicates of each sensor? Are there multiple measurement points or multiple eruptions?",
      "votes": null
    },
    {
      "id": "1075573",
      "postDate": "11/11/2020 19:35:06",
      "content": "<p>Hi Adam! Did you confirm the segment_ids to be overlapping?</p>",
      "rawMarkdown": "Hi Adam! Did you confirm the segment_ids to be overlapping?",
      "votes": null
    },
    {
      "id": "1089401",
      "postDate": "11/24/2020 13:20:58",
      "content": "<p>Have you resolved your question about <strong>data overlapping</strong>?<br>\nI'm still confused. All the data are from the <strong>same volcano</strong>, <strong>same measurement stations</strong>, <strong>same 10 sensors</strong>. But for example <code>segment_id</code>: 548932738 and 1049892705 have the same <code>time_to_eruption</code> of 23338589, but <strong>the data don't seem to be the same</strong>.</p>",
      "rawMarkdown": "Have you resolved your question about **data overlapping**?\nI'm still confused. All the data are from the **same volcano**, **same measurement stations**, **same 10 sensors**. But for example `segment_id`: 548932738 and 1049892705 have the same `time_to_eruption` of 23338589, but **the data don't seem to be the same**.",
      "votes": null
    },
    {
      "id": "1089601",
      "postDate": "11/24/2020 16:18:46",
      "content": "<p>multiple eruption sequences is the current theory</p>",
      "rawMarkdown": "multiple eruption sequences is the current theory",
      "votes": null
    },
    {
      "id": "1090484",
      "postDate": "11/25/2020 11:15:29",
      "content": "<p>Hi Adam.</p>\n<p><strong>multiple eruption!?</strong>  Really?  I thought it was the same eruption.<br>\nBoth the training set and the test set contain data from multiple eruptions. </p>\n<p>Lately, I read the Discussin page. It was already common sense that it was <strong>multiple eruption</strong>.</p>\n<p>In any case, thank you.</p>",
      "rawMarkdown": "Hi Adam.\n\n**multiple eruption!?**  Really?  I thought it was the same eruption.\nBoth the training set and the test set contain data from multiple eruptions. \n\nLately, I read the Discussin page. It was already common sense that it was **multiple eruption**.\n\nIn any case, thank you.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1047941,
      "author_name": "arunprathap",
      "author_url": "",
      "post_date": "10/13/2020 04:34:30",
      "content": "<p>At first look it seemed to me like a regression problem, but reading your thread made me realise that it more similar to a time sequence prediction problem. Only that instead of sequence we are to predict the time and expected sequence corresponding to volcanic eruption. Thanks.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1048200,
      "author_name": "flaxio",
      "author_url": "",
      "post_date": "10/13/2020 09:22:10",
      "content": "<p>Yes, you can assume the sampling rate at 100Hz and the time_to_eruption as the number of samples before the next eruption.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1066288,
          "author_name": "ajcostarino",
          "author_url": "",
          "post_date": "11/01/2020 15:30:31",
          "content": "<p>There are definitely overlapping segments right? Am I doing something wrong? If there are 4,000 segments each 60,000 centiseconds that means the <code>time_to_eruption</code> would have to span 240,000,000 centiseconds for no overlapping segments. Our competition only spans 40,000,000 centi seconds or 112 hours.</p>\n<p>And there are overlapping segments where sensors are on/off which means there are duplicates of each sensor? Are there multiple measurement points or multiple eruptions?</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1075573,
      "author_name": "chukalovskiy",
      "author_url": "",
      "post_date": "11/11/2020 19:35:06",
      "content": "<p>Hi Adam! Did you confirm the segment_ids to be overlapping?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1089401,
      "author_name": "amanooo",
      "author_url": "",
      "post_date": "11/24/2020 13:20:58",
      "content": "<p>Have you resolved your question about <strong>data overlapping</strong>?<br>\nI'm still confused. All the data are from the <strong>same volcano</strong>, <strong>same measurement stations</strong>, <strong>same 10 sensors</strong>. But for example <code>segment_id</code>: 548932738 and 1049892705 have the same <code>time_to_eruption</code> of 23338589, but <strong>the data don't seem to be the same</strong>.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1089601,
          "author_name": "ajcostarino",
          "author_url": "",
          "post_date": "11/24/2020 16:18:46",
          "content": "<p>multiple eruption sequences is the current theory</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1090484,
          "author_name": "amanooo",
          "author_url": "",
          "post_date": "11/25/2020 11:15:29",
          "content": "<p>Hi Adam.</p>\n<p><strong>multiple eruption!?</strong>  Really?  I thought it was the same eruption.<br>\nBoth the training set and the test set contain data from multiple eruptions. </p>\n<p>Lately, I read the Discussin page. It was already common sense that it was <strong>multiple eruption</strong>.</p>\n<p>In any case, thank you.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1047815": "Is time to eruption minutes, seconds, milliseconds, microseconds? How many time-steps between the measurements in the signal data?\n\nI cannot seem to find that metric anywhere in the documentation, if anybody has found it please inform!\n\nThanks!!\n\n**I see that it says that the observations are 10 min long, there are 60,000 rows per segment. 600 seconds per second. Meaning that the observations must be every 1/100 of a second. Let me know if I did that wrong.**\n\nCan we assume that the time to eruption is also in 1/100 of seconds (i.e centiseconds)",
    "1047941": "At first look it seemed to me like a regression problem, but reading your thread made me realise that it more similar to a time sequence prediction problem. Only that instead of sequence we are to predict the time and expected sequence corresponding to volcanic eruption. Thanks.",
    "1048200": "Yes, you can assume the sampling rate at 100Hz and the time_to_eruption as the number of samples before the next eruption.",
    "1066288": "There are definitely overlapping segments right? Am I doing something wrong? If there are 4,000 segments each 60,000 centiseconds that means the `time_to_eruption` would have to span 240,000,000 centiseconds for no overlapping segments. Our competition only spans 40,000,000 centi seconds or 112 hours.\n\n\nAnd there are overlapping segments where sensors are on/off which means there are duplicates of each sensor? Are there multiple measurement points or multiple eruptions?",
    "1075573": "Hi Adam! Did you confirm the segment_ids to be overlapping?",
    "1089401": "Have you resolved your question about **data overlapping**?\nI'm still confused. All the data are from the **same volcano**, **same measurement stations**, **same 10 sensors**. But for example `segment_id`: 548932738 and 1049892705 have the same `time_to_eruption` of 23338589, but **the data don't seem to be the same**.",
    "1089601": "multiple eruption sequences is the current theory",
    "1090484": "Hi Adam.\n\n**multiple eruption!?**  Really?  I thought it was the same eruption.\nBoth the training set and the test set contain data from multiple eruptions. \n\nLately, I read the Discussin page. It was already common sense that it was **multiple eruption**.\n\nIn any case, thank you."
  },
  "source": "meta"
}