{
  "id": 566606,
  "title": "[306th place] Bronze medal solution",
  "content": "<p>I was fortunate to barely make it into the bronze medal range in the first competition I participated in after Titanic.<br>\nI am very surprised that if the private score is correct, my ranking has improved by nearly 800 positions from the public score.<br>\nAs a memorandum for myself, I'll share what I have worked on.</p>\n<p>Here are my main solutions:</p>\n<ol>\n<li><p>Predicting the values after four types of target transformations using XGBoost, LightGBM, and CatBoost.</p></li>\n<li><p>Add the results from step 1 as features, and predict the values after the target transformation again using XGBoost and CatBoost.</p></li>\n<li><p>Create PyTorch NN Model.</p></li>\n<li><p>Ensemble steps 2 and 3.<br>\nI submitted the weighted sum of each prediction without converting to rank data. I did not use efs predictions by classifiers.</p></li>\n</ol>\n<p>Thanks to discussions and public notebook, I was able to gain a lot of new insights.<br>\nThank you, everyone.</p>",
  "comments": [
    {
      "id": 3142242,
      "author_name": "Vladimir Demidov",
      "author_url": "",
      "post_date": "2025-03-06T08:27:58.093000",
      "content": "<p>Nice jump. Interesting whether or not it could be better with rank data.</p>",
      "votes": 2,
      "replies": []
    },
    {
      "id": 3142408,
      "author_name": "Yahor Lahunovich",
      "author_url": "",
      "post_date": "2025-03-06T11:15:09.200000",
      "content": "<p>Congratulations!</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 3142245,
      "author_name": "",
      "author_url": "",
      "post_date": "2025-03-06T08:30:20.117000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "comment_count": 3,
  "messages": [
    {
      "id": 3142221,
      "postDate": "2025-03-06T08:07:50.887Z",
      "content": "<p>I was fortunate to barely make it into the bronze medal range in the first competition I participated in after Titanic.<br>\nI am very surprised that if the private score is correct, my ranking has improved by nearly 800 positions from the public score.<br>\nAs a memorandum for myself, I'll share what I have worked on.</p>\n<p>Here are my main solutions:</p>\n<ol>\n<li><p>Predicting the values after four types of target transformations using XGBoost, LightGBM, and CatBoost.</p></li>\n<li><p>Add the results from step 1 as features, and predict the values after the target transformation again using XGBoost and CatBoost.</p></li>\n<li><p>Create PyTorch NN Model.</p></li>\n<li><p>Ensemble steps 2 and 3.<br>\nI submitted the weighted sum of each prediction without converting to rank data. I did not use efs predictions by classifiers.</p></li>\n</ol>\n<p>Thanks to discussions and public notebook, I was able to gain a lot of new insights.<br>\nThank you, everyone.</p>",
      "rawMarkdown": "I was fortunate to barely make it into the bronze medal range in the first competition I participated in after Titanic.\nI am very surprised that if the private score is correct, my ranking has improved by nearly 800 positions from the public score.\nAs a memorandum for myself, I'll share what I have worked on.\n\nHere are my main solutions:\n\n1. Predicting the values after four types of target transformations using XGBoost, LightGBM, and CatBoost.\n\n2. Add the results from step 1 as features, and predict the values after the target transformation again using XGBoost and CatBoost.\n\n3. Create PyTorch NN Model.\n\n4. Ensemble steps 2 and 3.\nI submitted the weighted sum of each prediction without converting to rank data. I did not use efs predictions by classifiers.\n\nThanks to discussions and public notebook, I was able to gain a lot of new insights.\nThank you, everyone.",
      "votes": 6
    },
    {
      "id": 3142242,
      "postDate": "2025-03-06T08:27:58.093Z",
      "content": "<p>Nice jump. Interesting whether or not it could be better with rank data.</p>",
      "rawMarkdown": "Nice jump. Interesting whether or not it could be better with rank data.",
      "votes": 2
    },
    {
      "id": 3142408,
      "postDate": "2025-03-06T11:15:09.200Z",
      "content": "<p>Congratulations!</p>",
      "rawMarkdown": "Congratulations!"
    },
    {
      "id": 3142245,
      "postDate": "2025-03-06T08:30:20.117Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "complete": true,
  "source": "live",
  "fetched_at": "2026-10-19T19:47:46Z",
  "competition_id": 70942,
  "competition_slug": "equity-post-HCT-survival-predictions",
  "team_id": 13095260,
  "solution_url": "/competitions/equity-post-HCT-survival-predictions/writeups/miku-tsuchiya-306th-place-bronze-medal-solution"
}