{
  "id": 488204,
  "title": "The performance of XGBoost is far inferior to LightGBM",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/488204",
  "author_name": "wenjun zhang323",
  "post_date": "2024-04-01T14:43:31.482000",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I hope to train an XGBoost model to ensemble my LGB and CatBoost models. I used similar parameters as LGB, and the performance of XGBoost on the validation set is similar to LGB. However, when I submitted, the XGBoost LB score is much lower than LGB (about 0.02 lower). Has anyone encountered a similar situation? I'm not sure if there's an error in my code, but the function for predicting the test set and the validation set is clearly the same.</p>",
  "messages": [
    {
      "id": 2727031,
      "postDate": "2024-04-01T14:43:31.483Z",
      "content": "<p>I hope to train an XGBoost model to ensemble my LGB and CatBoost models. I used similar parameters as LGB, and the performance of XGBoost on the validation set is similar to LGB. However, when I submitted, the XGBoost LB score is much lower than LGB (about 0.02 lower). Has anyone encountered a similar situation? I'm not sure if there's an error in my code, but the function for predicting the test set and the validation set is clearly the same.</p>",
      "rawMarkdown": "I hope to train an XGBoost model to ensemble my LGB and CatBoost models. I used similar parameters as LGB, and the performance of XGBoost on the validation set is similar to LGB. However, when I submitted, the XGBoost LB score is much lower than LGB (about 0.02 lower). Has anyone encountered a similar situation? I'm not sure if there's an error in my code, but the function for predicting the test set and the validation set is clearly the same.",
      "votes": 2
    },
    {
      "id": 2737645,
      "postDate": "2024-04-05T21:11:20.267Z",
      "content": "<p>Yes, I have similar experience with xgboost. </p>",
      "rawMarkdown": "Yes, I have similar experience with xgboost. "
    },
    {
      "id": 2738454,
      "postDate": "2024-04-06T10:52:08.077Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2737645,
      "author_name": "Rafał Pawłowski",
      "author_url": "",
      "post_date": "2024-04-05T21:11:20.267000",
      "content": "<p>Yes, I have similar experience with xgboost. </p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 2738454,
      "author_name": "",
      "author_url": "",
      "post_date": "2024-04-06T10:52:08.077000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2727031": "I hope to train an XGBoost model to ensemble my LGB and CatBoost models. I used similar parameters as LGB, and the performance of XGBoost on the validation set is similar to LGB. However, when I submitted, the XGBoost LB score is much lower than LGB (about 0.02 lower). Has anyone encountered a similar situation? I'm not sure if there's an error in my code, but the function for predicting the test set and the validation set is clearly the same.",
    "2737645": "Yes, I have similar experience with xgboost. ",
    "2738454": ""
  }
}