{
  "id": 451510,
  "title": "Target Encoders - LeaveOneOutEncoder - top1 ",
  "url": "/competitions/open-problems-single-cell-perturbations/discussion/451510",
  "author_name": "Alexander Chervov",
  "post_date": "2023-10-29T09:59:21.966000",
  "votes": 5,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Here are some benchmarks of  several target encoders from the package \"<a href=\"https://contrib.scikit-learn.org/category_encoders/\" target=\"_blank\">Category Encoders</a>\".</p>\n<p>Surprisingly the LeaveOneOutEncoder seems to be top1 for the most cases (at least for default params).<br>\nThe setup considered: tsvd30+encoder+Ridge+tsvd.inverse_transform . <br>\nRidge regularization alpha was carefully optimized for each case. </p>\n<p><a href=\"https://www.kaggle.com/code/alexandervc/op2-target-encoders\" target=\"_blank\">See the notebook</a>. LB score for submit is 0.608,  everybody is welcome to check  others encoders submits  saved in \"output tab\" of the notebook. </p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2Fcf1817bf7dd9bd1459ca2ddc7084e463%2FScreenshot%202023-10-29%20104838.png?generation=1698572992458641&amp;alt=media\" alt=\"\"></p>",
  "messages": [
    {
      "id": 2503630,
      "postDate": "2023-10-29T09:59:21.967Z",
      "content": "<p>Here are some benchmarks of  several target encoders from the package \"<a href=\"https://contrib.scikit-learn.org/category_encoders/\" target=\"_blank\">Category Encoders</a>\".</p>\n<p>Surprisingly the LeaveOneOutEncoder seems to be top1 for the most cases (at least for default params).<br>\nThe setup considered: tsvd30+encoder+Ridge+tsvd.inverse_transform . <br>\nRidge regularization alpha was carefully optimized for each case. </p>\n<p><a href=\"https://www.kaggle.com/code/alexandervc/op2-target-encoders\" target=\"_blank\">See the notebook</a>. LB score for submit is 0.608,  everybody is welcome to check  others encoders submits  saved in \"output tab\" of the notebook. </p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2Fcf1817bf7dd9bd1459ca2ddc7084e463%2FScreenshot%202023-10-29%20104838.png?generation=1698572992458641&amp;alt=media\" alt=\"\"></p>",
      "rawMarkdown": "Here are some benchmarks of  several target encoders from the package \"[Category Encoders](https://contrib.scikit-learn.org/category_encoders/)\".\n\nSurprisingly the LeaveOneOutEncoder seems to be top1 for the most cases (at least for default params).\nThe setup considered: tsvd30+encoder+Ridge+tsvd.inverse_transform . \nRidge regularization alpha was carefully optimized for each case. \n\n[See the notebook](https://www.kaggle.com/code/alexandervc/op2-target-encoders). LB score for submit is 0.608,  everybody is welcome to check  others encoders submits  saved in \"output tab\" of the notebook. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2Fcf1817bf7dd9bd1459ca2ddc7084e463%2FScreenshot%202023-10-29%20104838.png?generation=1698572992458641&alt=media)",
      "votes": 5
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2503630": "Here are some benchmarks of  several target encoders from the package \"[Category Encoders](https://contrib.scikit-learn.org/category_encoders/)\".\n\nSurprisingly the LeaveOneOutEncoder seems to be top1 for the most cases (at least for default params).\nThe setup considered: tsvd30+encoder+Ridge+tsvd.inverse_transform . \nRidge regularization alpha was carefully optimized for each case. \n\n[See the notebook](https://www.kaggle.com/code/alexandervc/op2-target-encoders). LB score for submit is 0.608,  everybody is welcome to check  others encoders submits  saved in \"output tab\" of the notebook. \n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2262596%2Fcf1817bf7dd9bd1459ca2ddc7084e463%2FScreenshot%202023-10-29%20104838.png?generation=1698572992458641&alt=media)"
  }
}