{
  "id": 244750,
  "title": "What to do about the Xiaomi Mi8 phone?",
  "url": "/competitions/google-smartphone-decimeter-challenge/discussion/244750",
  "author_name": "",
  "post_date": "2021-06-08T06:54:17.552103900Z",
  "votes": 10,
  "comment_count": 1,
  "views": 0,
  "content": "<p>It seem that by far worst offender in the dataset is the Xiaomi <strong>Mi8</strong>  phone, whose position is sometimes off by as far as 8km (5 miles!). This is almost certainly due to it having on occasions a <code>timeSinceFirstFixSeconds</code> (the time since the first location fix) of over  half an hour, and having some crazy heights, both above and underground!</p>\n<p>I suspect that doing something about this 'prodigal' phone will result in substantial improvements overall, as most of the other phones are <em>much</em> more well behaved!</p>",
  "messages": [
    {
      "id": "1340638",
      "postDate": "06/08/2021 06:54:17",
      "content": "<p>It seem that by far worst offender in the dataset is the Xiaomi <strong>Mi8</strong>  phone, whose position is sometimes off by as far as 8km (5 miles!). This is almost certainly due to it having on occasions a <code>timeSinceFirstFixSeconds</code> (the time since the first location fix) of over  half an hour, and having some crazy heights, both above and underground!</p>\n<p>I suspect that doing something about this 'prodigal' phone will result in substantial improvements overall, as most of the other phones are <em>much</em> more well behaved!</p>",
      "rawMarkdown": "It seem that by far worst offender in the dataset is the Xiaomi **Mi8**  phone, whose position is sometimes off by as far as 8km (5 miles!). This is almost certainly due to it having on occasions a `timeSinceFirstFixSeconds` (the time since the first location fix) of over  half an hour, and having some crazy heights, both above and underground!\n\nI suspect that doing something about this 'prodigal' phone will result in substantial improvements overall, as most of the other phones are *much* more well behaved!",
      "votes": null
    },
    {
      "id": "1341191",
      "postDate": "06/08/2021 13:46:00",
      "content": "<p>By replacing the outliers with an interpolated value the mean and std error of mi8 decreases a lot:<br>\n<a href=\"https://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction\" target=\"_blank\">https://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction</a></p>\n<pre><code>mean                     std                       max          \n3.7588691575223705      86.65617078472657   8231.584396457476   \n</code></pre>\n<p>becomes</p>\n<pre><code>2.485252958240607      13.04238722687981   1256.2067156207888\n</code></pre>\n<p>The kalman smoothing brings it further down to 500 meters max error</p>\n<pre><code>2.347146450969712      8.795083379555939   508.5690008061522\n</code></pre>\n<p>Worst phone seems to be SamsungS20Ultra even after Outlier Correction and Kalman</p>\n<pre><code>7.097146931844713      7.506729858076974   100.27818665405384\n</code></pre>",
      "rawMarkdown": "By replacing the outliers with an interpolated value the mean and std error of mi8 decreases a lot:\nhttps://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction\n```\nmean\t\t\t         std\t\t\t           max\t\t\t\n3.7588691575223705 \t 86.65617078472657 \t 8231.584396457476 \t \n```\nbecomes\n```\n2.485252958240607 \t 13.04238722687981 \t 1256.2067156207888\n```\n\nThe kalman smoothing brings it further down to 500 meters max error\n```\n2.347146450969712 \t 8.795083379555939 \t 508.5690008061522\n```\n\nWorst phone seems to be SamsungS20Ultra even after Outlier Correction and Kalman\n```\n7.097146931844713 \t 7.506729858076974 \t 100.27818665405384\n```",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1341191,
      "author_name": "marekwyborski",
      "author_url": "",
      "post_date": "06/08/2021 13:46:00",
      "content": "<p>By replacing the outliers with an interpolated value the mean and std error of mi8 decreases a lot:<br>\n<a href=\"https://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction\" target=\"_blank\">https://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction</a></p>\n<pre><code>mean                     std                       max          \n3.7588691575223705      86.65617078472657   8231.584396457476   \n</code></pre>\n<p>becomes</p>\n<pre><code>2.485252958240607      13.04238722687981   1256.2067156207888\n</code></pre>\n<p>The kalman smoothing brings it further down to 500 meters max error</p>\n<pre><code>2.347146450969712      8.795083379555939   508.5690008061522\n</code></pre>\n<p>Worst phone seems to be SamsungS20Ultra even after Outlier Correction and Kalman</p>\n<pre><code>7.097146931844713      7.506729858076974   100.27818665405384\n</code></pre>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1340638": "It seem that by far worst offender in the dataset is the Xiaomi **Mi8**  phone, whose position is sometimes off by as far as 8km (5 miles!). This is almost certainly due to it having on occasions a `timeSinceFirstFixSeconds` (the time since the first location fix) of over  half an hour, and having some crazy heights, both above and underground!\n\nI suspect that doing something about this 'prodigal' phone will result in substantial improvements overall, as most of the other phones are *much* more well behaved!",
    "1341191": "By replacing the outliers with an interpolated value the mean and std error of mi8 decreases a lot:\nhttps://www.kaggle.com/dehokanta/baseline-post-processing-by-outlier-correction\n```\nmean\t\t\t         std\t\t\t           max\t\t\t\n3.7588691575223705 \t 86.65617078472657 \t 8231.584396457476 \t \n```\nbecomes\n```\n2.485252958240607 \t 13.04238722687981 \t 1256.2067156207888\n```\n\nThe kalman smoothing brings it further down to 500 meters max error\n```\n2.347146450969712 \t 8.795083379555939 \t 508.5690008061522\n```\n\nWorst phone seems to be SamsungS20Ultra even after Outlier Correction and Kalman\n```\n7.097146931844713 \t 7.506729858076974 \t 100.27818665405384\n```"
  },
  "source": "meta"
}