{
  "id": 462797,
  "title": "152th solution",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/462797",
  "author_name": "RyotaroAbe",
  "post_date": "2023-12-21T15:37:53.563000",
  "votes": 6,
  "comment_count": 0,
  "views": 0,
  "content": "<h1>Overview</h1>\n<p>First, I would like to express my sincere thanks to the organizers of the Child Mind Institute - Detect Sleep States Competition.<br>\nI tried this competition based on <a href=\"https://www.kaggle.com/tubotubo\" target=\"_blank\">@tubotubo</a> 's baseline. I extend my gratitude.</p>\n<h1>My Solution</h1>\n<ul>\n<li>Changed the way of validation<ul>\n<li>Divided into groups using GroupKFold, ensuring an equal distribution of number of events.</li>\n<li>0.727</li></ul></li>\n<li>Clipped enmo<ul>\n<li>Clipped at 1 as outliers were significant</li>\n<li>0.724</li></ul></li>\n<li>Seed averaging with three different values<ul>\n<li>0: 0.725</li>\n<li>21: 0.724</li>\n<li>42: 0.723</li></ul></li>\n<li>Ensamble models<ul>\n<li>Averaged the predictions of each model, and then applied post-processing using find_peaks.</li>\n<li>0.73~0.735</li></ul></li>\n</ul>\n<h1>Did Not Work</h1>\n<ul>\n<li>window for input was expanded from 8 hours to 24 hours<ul>\n<li>I wanted to include information from the previous day and other times in the models.</li></ul></li>\n<li>Changing parameters of Gaussian label<ul>\n<li>Attempted to pass smoother labels to the model, but did not see an improvement in accuracy.</li></ul></li>\n<li>Changing parameters of find_peaks</li>\n<li>Deleting low score predictions in post-process<ul>\n<li>Set a threshold for scores, but it was not successful.</li></ul></li>\n</ul>",
  "messages": [
    {
      "id": 2569834,
      "postDate": "2023-12-21T15:37:53.563Z",
      "content": "<h1>Overview</h1>\n<p>First, I would like to express my sincere thanks to the organizers of the Child Mind Institute - Detect Sleep States Competition.<br>\nI tried this competition based on <a href=\"https://www.kaggle.com/tubotubo\" target=\"_blank\">@tubotubo</a> 's baseline. I extend my gratitude.</p>\n<h1>My Solution</h1>\n<ul>\n<li>Changed the way of validation<ul>\n<li>Divided into groups using GroupKFold, ensuring an equal distribution of number of events.</li>\n<li>0.727</li></ul></li>\n<li>Clipped enmo<ul>\n<li>Clipped at 1 as outliers were significant</li>\n<li>0.724</li></ul></li>\n<li>Seed averaging with three different values<ul>\n<li>0: 0.725</li>\n<li>21: 0.724</li>\n<li>42: 0.723</li></ul></li>\n<li>Ensamble models<ul>\n<li>Averaged the predictions of each model, and then applied post-processing using find_peaks.</li>\n<li>0.73~0.735</li></ul></li>\n</ul>\n<h1>Did Not Work</h1>\n<ul>\n<li>window for input was expanded from 8 hours to 24 hours<ul>\n<li>I wanted to include information from the previous day and other times in the models.</li></ul></li>\n<li>Changing parameters of Gaussian label<ul>\n<li>Attempted to pass smoother labels to the model, but did not see an improvement in accuracy.</li></ul></li>\n<li>Changing parameters of find_peaks</li>\n<li>Deleting low score predictions in post-process<ul>\n<li>Set a threshold for scores, but it was not successful.</li></ul></li>\n</ul>",
      "rawMarkdown": "# Overview\nFirst, I would like to express my sincere thanks to the organizers of the Child Mind Institute - Detect Sleep States Competition.\nI tried this competition based on @tubotubo 's baseline. I extend my gratitude.\n\n# My Solution\n- Changed the way of validation\n  - Divided into groups using GroupKFold, ensuring an equal distribution of number of events.\n  - 0.727\n- Clipped enmo\n  - Clipped at 1 as outliers were significant\n  - 0.724\n- Seed averaging with three different values\n  - 0: 0.725\n  - 21: 0.724\n  - 42: 0.723\n- Ensamble models\n  - Averaged the predictions of each model, and then applied post-processing using find_peaks.\n  - 0.73~0.735\n\n# Did Not Work\n- window for input was expanded from 8 hours to 24 hours\n  - I wanted to include information from the previous day and other times in the models.\n- Changing parameters of Gaussian label\n  - Attempted to pass smoother labels to the model, but did not see an improvement in accuracy.\n- Changing parameters of find_peaks\n- Deleting low score predictions in post-process\n  - Set a threshold for scores, but it was not successful.",
      "votes": 6
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2569834": "# Overview\nFirst, I would like to express my sincere thanks to the organizers of the Child Mind Institute - Detect Sleep States Competition.\nI tried this competition based on @tubotubo 's baseline. I extend my gratitude.\n\n# My Solution\n- Changed the way of validation\n  - Divided into groups using GroupKFold, ensuring an equal distribution of number of events.\n  - 0.727\n- Clipped enmo\n  - Clipped at 1 as outliers were significant\n  - 0.724\n- Seed averaging with three different values\n  - 0: 0.725\n  - 21: 0.724\n  - 42: 0.723\n- Ensamble models\n  - Averaged the predictions of each model, and then applied post-processing using find_peaks.\n  - 0.73~0.735\n\n# Did Not Work\n- window for input was expanded from 8 hours to 24 hours\n  - I wanted to include information from the previous day and other times in the models.\n- Changing parameters of Gaussian label\n  - Attempted to pass smoother labels to the model, but did not see an improvement in accuracy.\n- Changing parameters of find_peaks\n- Deleting low score predictions in post-process\n  - Set a threshold for scores, but it was not successful."
  }
}