{
  "id": 439039,
  "title": "What data should we use for training? ",
  "url": "/competitions/child-mind-institute-detect-sleep-states/discussion/439039",
  "author_name": "Manu",
  "post_date": "2023-09-13T13:27:52.452000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I have seen several notebooks that rule out every series with any null-step event. This way the number of different series used for training is reduced a lot.  I think it could be a better solution to processed data in a similar way to this notebook: <a href=\"https://www.kaggle.com/code/mpp1998/eda-zzz\" target=\"_blank\">mpp1998/eda-zzz</a> so not the full series is ruled out but only the parts which are unlabelled. I identify unlabelled rows if I don´t kown when was its previous/next event (because there are null-step events between them).<br>\n <img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13599107%2F39886eb4aadc2d448b73a2ada04243cc%2FCaptura.PNG?generation=1694611406450158&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": 2436288,
      "postDate": "2023-09-13T13:27:52.453Z",
      "content": "<p>I have seen several notebooks that rule out every series with any null-step event. This way the number of different series used for training is reduced a lot.  I think it could be a better solution to processed data in a similar way to this notebook: <a href=\"https://www.kaggle.com/code/mpp1998/eda-zzz\" target=\"_blank\">mpp1998/eda-zzz</a> so not the full series is ruled out but only the parts which are unlabelled. I identify unlabelled rows if I don´t kown when was its previous/next event (because there are null-step events between them).<br>\n <img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13599107%2F39886eb4aadc2d448b73a2ada04243cc%2FCaptura.PNG?generation=1694611406450158&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "I have seen several notebooks that rule out every series with any null-step event. This way the number of different series used for training is reduced a lot.  I think it could be a better solution to processed data in a similar way to this notebook: [mpp1998/eda-zzz](https://www.kaggle.com/code/mpp1998/eda-zzz) so not the full series is ruled out but only the parts which are unlabelled. I identify unlabelled rows if I don´t kown when was its previous/next event (because there are null-step events between them).\n ![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13599107%2F39886eb4aadc2d448b73a2ada04243cc%2FCaptura.PNG?generation=1694611406450158&alt=media)",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2436288": "I have seen several notebooks that rule out every series with any null-step event. This way the number of different series used for training is reduced a lot.  I think it could be a better solution to processed data in a similar way to this notebook: [mpp1998/eda-zzz](https://www.kaggle.com/code/mpp1998/eda-zzz) so not the full series is ruled out but only the parts which are unlabelled. I identify unlabelled rows if I don´t kown when was its previous/next event (because there are null-step events between them).\n ![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F13599107%2F39886eb4aadc2d448b73a2ada04243cc%2FCaptura.PNG?generation=1694611406450158&alt=media)"
  }
}