{
  "id": 80597,
  "title": "Data Direction",
  "url": "/competitions/LANL-Earthquake-Prediction/discussion/80597",
  "author_name": "",
  "post_date": "2019-02-14T17:25:47.699145400Z",
  "votes": 2,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hello,\nwhat is the data direction. Is it top to bottom or bottom to top?</p>\n\n<p>In other words if I take iloc [-100:-1] shall I predict value -1 or -100? </p>\n\n<p>Based on the 'time_to_failure' column, I assume it should be bottom to top (predict -100 for set [-100:-1])</p>\n\n<p>I would appreciate clear statement on that.</p>\n\n<p>Regards</p>",
  "messages": [
    {
      "id": "471628",
      "postDate": "02/14/2019 17:25:47",
      "content": "<p>Hello,\nwhat is the data direction. Is it top to bottom or bottom to top?</p>\n\n<p>In other words if I take iloc [-100:-1] shall I predict value -1 or -100? </p>\n\n<p>Based on the 'time_to_failure' column, I assume it should be bottom to top (predict -100 for set [-100:-1])</p>\n\n<p>I would appreciate clear statement on that.</p>\n\n<p>Regards</p>",
      "rawMarkdown": "Hello,\nwhat is the data direction. Is it top to bottom or bottom to top?\n\nIn other words if I take iloc [-100:-1] shall I predict value -1 or -100? \n\nBased on the 'time\\_to\\_failure' column, I assume it should be bottom to top (predict -100 for set [-100:-1])\n\nI would appreciate clear statement on that.\n\nRegards",
      "votes": null
    },
    {
      "id": "471670",
      "postDate": "02/14/2019 18:40:20",
      "content": "<p>I have no idea what you are asking.</p>",
      "rawMarkdown": "I have no idea what you are asking.",
      "votes": null
    },
    {
      "id": "471824",
      "postDate": "02/15/2019 00:05:37",
      "content": "<p>For each group of acoustic_data values, we have to predict time_to_failure that corresponds to the last acoustic_data value. In your example, you have to predict the time_to_failure value at -1. As time goes the number of acoustic_data points increase and the remaining time (time_to_failure values) to an labquake decrease. </p>",
      "rawMarkdown": "For each group of acoustic_data values, we have to predict time_to_failure that corresponds to the last acoustic_data value. In your example, you have to predict the time_to_failure value at -1. As time goes the number of acoustic_data points increase and the remaining time (time_to_failure values) to an labquake decrease.",
      "votes": null
    },
    {
      "id": "472039",
      "postDate": "02/15/2019 08:51:05",
      "content": "<p>If I got the experiment right the time to failure should decrease with every step as we are getting closure to the event of failure. In the same time the time to failure in the training set is increasing with every step. So the question is what is the direction of the time series.</p>",
      "rawMarkdown": "If I got the experiment right the time to failure should decrease with every step as we are getting closure to the event of failure. In the same time the time to failure in the training set is increasing with every step. So the question is what is the direction of the time series.",
      "votes": null
    },
    {
      "id": "472065",
      "postDate": "02/15/2019 09:31:09",
      "content": "<p>&gt; In the same time the time to failure in the training set is increasing with every step.</p>\n\n<p>No, the values in the column <code>time_to_failure</code> of the file <code>train.csv</code> are decreasing, except for the positions where a quake occured and a new experiment begins.</p>",
      "rawMarkdown": "&gt; In the same time the time to failure in the training set is increasing with every step.\n\nNo, the values in the column `time_to_failure` of the file `train.csv` are decreasing, except for the positions where a quake occured and a new experiment begins.",
      "votes": null
    },
    {
      "id": "472072",
      "postDate": "02/15/2019 09:42:29",
      "content": "<p>Maybe the way you extract ttf values from the train file orders them backwards. Try to read them from the small row indeces to the larger.  </p>",
      "rawMarkdown": "Maybe the way you extract ttf values from the train file orders them backwards. Try to read them from the small row indeces to the larger.",
      "votes": null
    },
    {
      "id": "472075",
      "postDate": "02/15/2019 09:50:30",
      "content": "<p>My fault, thanks and apologies.</p>",
      "rawMarkdown": "My fault, thanks and apologies.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 471670,
      "author_name": "danjel",
      "author_url": "",
      "post_date": "02/14/2019 18:40:20",
      "content": "<p>I have no idea what you are asking.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 471824,
      "author_name": "petya5q",
      "author_url": "",
      "post_date": "02/15/2019 00:05:37",
      "content": "<p>For each group of acoustic_data values, we have to predict time_to_failure that corresponds to the last acoustic_data value. In your example, you have to predict the time_to_failure value at -1. As time goes the number of acoustic_data points increase and the remaining time (time_to_failure values) to an labquake decrease. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 472039,
      "author_name": "kovachev",
      "author_url": "",
      "post_date": "02/15/2019 08:51:05",
      "content": "<p>If I got the experiment right the time to failure should decrease with every step as we are getting closure to the event of failure. In the same time the time to failure in the training set is increasing with every step. So the question is what is the direction of the time series.</p>",
      "votes": null,
      "replies": [
        {
          "id": 472065,
          "author_name": "danjel",
          "author_url": "",
          "post_date": "02/15/2019 09:31:09",
          "content": "<p>&gt; In the same time the time to failure in the training set is increasing with every step.</p>\n\n<p>No, the values in the column <code>time_to_failure</code> of the file <code>train.csv</code> are decreasing, except for the positions where a quake occured and a new experiment begins.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 472072,
          "author_name": "petya5q",
          "author_url": "",
          "post_date": "02/15/2019 09:42:29",
          "content": "<p>Maybe the way you extract ttf values from the train file orders them backwards. Try to read them from the small row indeces to the larger.  </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 472075,
          "author_name": "kovachev",
          "author_url": "",
          "post_date": "02/15/2019 09:50:30",
          "content": "<p>My fault, thanks and apologies.</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "471628": "Hello,\nwhat is the data direction. Is it top to bottom or bottom to top?\n\nIn other words if I take iloc [-100:-1] shall I predict value -1 or -100? \n\nBased on the 'time\\_to\\_failure' column, I assume it should be bottom to top (predict -100 for set [-100:-1])\n\nI would appreciate clear statement on that.\n\nRegards",
    "471670": "I have no idea what you are asking.",
    "471824": "For each group of acoustic_data values, we have to predict time_to_failure that corresponds to the last acoustic_data value. In your example, you have to predict the time_to_failure value at -1. As time goes the number of acoustic_data points increase and the remaining time (time_to_failure values) to an labquake decrease.",
    "472039": "If I got the experiment right the time to failure should decrease with every step as we are getting closure to the event of failure. In the same time the time to failure in the training set is increasing with every step. So the question is what is the direction of the time series.",
    "472065": "&gt; In the same time the time to failure in the training set is increasing with every step.\n\nNo, the values in the column `time_to_failure` of the file `train.csv` are decreasing, except for the positions where a quake occured and a new experiment begins.",
    "472072": "Maybe the way you extract ttf values from the train file orders them backwards. Try to read them from the small row indeces to the larger.",
    "472075": "My fault, thanks and apologies."
  },
  "source": "meta"
}