{
  "id": 374502,
  "title": "My LGBMRanker performs poorly ",
  "url": "/competitions/otto-recommender-system/discussion/374502",
  "author_name": "Rayan-aay",
  "post_date": "2022-12-27T16:12:58.631000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hello community, I'm using LGBMRanker to select candidates from each user among 50  strong candidates for each from covisitation-matrix. </p>\n<p>I was expecting, in the worst case, a little boosting in the validation score. However, the hiits decreased, knowing:</p>\n<ul>\n<li>The features used are well extracted.</li>\n<li>The NDCG@20 is increasing at each iteration.</li>\n</ul>\n<p>PS : I used a 64 GB Ram computer with 8 CPU Cores</p>",
  "messages": [
    {
      "id": 2077485,
      "postDate": "2022-12-27T16:12:58.630Z",
      "content": "<p>Hello community, I'm using LGBMRanker to select candidates from each user among 50  strong candidates for each from covisitation-matrix. </p>\n<p>I was expecting, in the worst case, a little boosting in the validation score. However, the hiits decreased, knowing:</p>\n<ul>\n<li>The features used are well extracted.</li>\n<li>The NDCG@20 is increasing at each iteration.</li>\n</ul>\n<p>PS : I used a 64 GB Ram computer with 8 CPU Cores</p>",
      "rawMarkdown": "Hello community, I'm using LGBMRanker to select candidates from each user among 50  strong candidates for each from covisitation-matrix. \n\nI was expecting, in the worst case, a little boosting in the validation score. However, the hiits decreased, knowing:\n\n- The features used are well extracted.\n- The NDCG@20 is increasing at each iteration.\n\nPS : I used a 64 GB Ram computer with 8 CPU Cores",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2077485": "Hello community, I'm using LGBMRanker to select candidates from each user among 50  strong candidates for each from covisitation-matrix. \n\nI was expecting, in the worst case, a little boosting in the validation score. However, the hiits decreased, knowing:\n\n- The features used are well extracted.\n- The NDCG@20 is increasing at each iteration.\n\nPS : I used a 64 GB Ram computer with 8 CPU Cores"
  }
}