{
  "id": 215346,
  "title": "How to predict the start position?",
  "url": "/competitions/indoor-location-navigation/discussion/215346",
  "author_name": "",
  "post_date": "2021-01-29T13:59:12.747698800Z",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>I think this competition is divided the two problem.</p>\n<ol>\n<li>Predict the start position</li>\n<li>Predict the next position using the previous position (in other words, predict the path from the previous position.)</li>\n</ol>\n<p>The acceleration of the device seems to be useful for the second problem, but the first problem is very difficult to solve.</p>\n<p>Do you have any idea?</p>",
  "messages": [
    {
      "id": "1176093",
      "postDate": "01/29/2021 13:59:12",
      "content": "<p>I think this competition is divided the two problem.</p>\n<ol>\n<li>Predict the start position</li>\n<li>Predict the next position using the previous position (in other words, predict the path from the previous position.)</li>\n</ol>\n<p>The acceleration of the device seems to be useful for the second problem, but the first problem is very difficult to solve.</p>\n<p>Do you have any idea?</p>",
      "rawMarkdown": "I think this competition is divided the two problem.\n1. Predict the start position\n2. Predict the next position using the previous position (in other words, predict the path from the previous position.)\n\nThe acceleration of the device seems to be useful for the second problem, but the first problem is very difficult to solve.\n\nDo you have any idea?",
      "votes": null
    },
    {
      "id": "1178136",
      "postDate": "01/30/2021 17:15:31",
      "content": "<p>I haven't checked all the path trace files but the time differences between the first sensor data and the start position do not seem to be more than a few seconds. I think the solution is to use the data within a deltaT of the start position's timestamp.</p>",
      "rawMarkdown": "I haven't checked all the path trace files but the time differences between the first sensor data and the start position do not seem to be more than a few seconds. I think the solution is to use the data within a deltaT of the start position's timestamp.",
      "votes": null
    },
    {
      "id": "1180834",
      "postDate": "02/01/2021 13:48:37",
      "content": "<p>Thanks, I overlooked that there is no problem with using a little future data :)</p>",
      "rawMarkdown": "Thanks, I overlooked that there is no problem with using a little future data :)",
      "votes": null
    },
    {
      "id": "1242332",
      "postDate": "03/17/2021 14:51:10",
      "content": "<p>Hi there! Just wondering, because we are not provided with any initial position at all in the test set, so how are we going to determine the start position in the test set?</p>",
      "rawMarkdown": "Hi there! Just wondering, because we are not provided with any initial position at all in the test set, so how are we going to determine the start position in the test set?",
      "votes": null
    },
    {
      "id": "1255731",
      "postDate": "03/29/2021 06:26:45",
      "content": "<p>Hi <a href=\"https://www.kaggle.com/angeltang6918\" target=\"_blank\">@angeltang6918</a>, as <a href=\"https://www.kaggle.com/tolgadincer\" target=\"_blank\">@tolgadincer</a> pointed out, we can look at information after the timestamp too and find the start position in reverse direction</p>",
      "rawMarkdown": "Hi @angeltang6918, as @tolgadincer pointed out, we can look at information after the timestamp too and find the start position in reverse direction",
      "votes": null
    },
    {
      "id": "1255847",
      "postDate": "03/29/2021 09:17:24",
      "content": "<p>Thanks for your reply!</p>",
      "rawMarkdown": "Thanks for your reply!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1178136,
      "author_name": "tolgadincer",
      "author_url": "",
      "post_date": "01/30/2021 17:15:31",
      "content": "<p>I haven't checked all the path trace files but the time differences between the first sensor data and the start position do not seem to be more than a few seconds. I think the solution is to use the data within a deltaT of the start position's timestamp.</p>",
      "votes": null,
      "replies": [
        {
          "id": 1180834,
          "author_name": "tj0612",
          "author_url": "",
          "post_date": "02/01/2021 13:48:37",
          "content": "<p>Thanks, I overlooked that there is no problem with using a little future data :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1242332,
      "author_name": "angeltang6918",
      "author_url": "",
      "post_date": "03/17/2021 14:51:10",
      "content": "<p>Hi there! Just wondering, because we are not provided with any initial position at all in the test set, so how are we going to determine the start position in the test set?</p>",
      "votes": null,
      "replies": [
        {
          "id": 1255731,
          "author_name": "suryajrrafl",
          "author_url": "",
          "post_date": "03/29/2021 06:26:45",
          "content": "<p>Hi <a href=\"https://www.kaggle.com/angeltang6918\" target=\"_blank\">@angeltang6918</a>, as <a href=\"https://www.kaggle.com/tolgadincer\" target=\"_blank\">@tolgadincer</a> pointed out, we can look at information after the timestamp too and find the start position in reverse direction</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1255847,
          "author_name": "angeltang6918",
          "author_url": "",
          "post_date": "03/29/2021 09:17:24",
          "content": "<p>Thanks for your reply!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1176093": "I think this competition is divided the two problem.\n1. Predict the start position\n2. Predict the next position using the previous position (in other words, predict the path from the previous position.)\n\nThe acceleration of the device seems to be useful for the second problem, but the first problem is very difficult to solve.\n\nDo you have any idea?",
    "1178136": "I haven't checked all the path trace files but the time differences between the first sensor data and the start position do not seem to be more than a few seconds. I think the solution is to use the data within a deltaT of the start position's timestamp.",
    "1180834": "Thanks, I overlooked that there is no problem with using a little future data :)",
    "1242332": "Hi there! Just wondering, because we are not provided with any initial position at all in the test set, so how are we going to determine the start position in the test set?",
    "1255731": "Hi @angeltang6918, as @tolgadincer pointed out, we can look at information after the timestamp too and find the start position in reverse direction",
    "1255847": "Thanks for your reply!"
  },
  "source": "meta"
}