{
  "id": 392156,
  "title": "Stuck 😓 (Help Needed)",
  "url": "/competitions/predict-student-performance-from-game-play/discussion/392156",
  "author_name": "",
  "post_date": "2023-03-03T20:55:44.480613Z",
  "votes": 4,
  "comment_count": 5,
  "views": 0,
  "content": "<p>Hii,<br>\nAfter adding about 80-90 features to the model, a point comes where the score decreases with any new bunch of features added. I am not able to increase the score any further, no matter what I do.<br>\nIs this because the new features are just not helpful, there is too much noise in the previous features or there is some other reason?<br>\nI even tried using alpha parameter of xgboost, but it doesn't help. 😓<br>\nWhat can I do to fix this? <br>\nThanks.</p>",
  "messages": [
    {
      "id": "2168062",
      "postDate": "03/03/2023 20:55:44",
      "content": "<p>Hii,<br>\nAfter adding about 80-90 features to the model, a point comes where the score decreases with any new bunch of features added. I am not able to increase the score any further, no matter what I do.<br>\nIs this because the new features are just not helpful, there is too much noise in the previous features or there is some other reason?<br>\nI even tried using alpha parameter of xgboost, but it doesn't help. 😓<br>\nWhat can I do to fix this? <br>\nThanks.</p>",
      "rawMarkdown": "Hii,\nAfter adding about 80-90 features to the model, a point comes where the score decreases with any new bunch of features added. I am not able to increase the score any further, no matter what I do.\nIs this because the new features are just not helpful, there is too much noise in the previous features or there is some other reason?\nI even tried using alpha parameter of xgboost, but it doesn't help. 😓\nWhat can I do to fix this? \nThanks.",
      "votes": null
    },
    {
      "id": "2168063",
      "postDate": "03/03/2023 20:59:39",
      "content": "<p>I think your features are not good. For me never happens if I use alpha=8</p>",
      "rawMarkdown": "I think your features are not good. For me never happens if I use alpha=8",
      "votes": null
    },
    {
      "id": "2168069",
      "postDate": "03/03/2023 21:03:49",
      "content": "<p>Yeah maybe. I'll check. Can this also happen if the newly added features are very similar to the previous features?<br>\nThanks.</p>",
      "rawMarkdown": "Yeah maybe. I'll check. Can this also happen if the newly added features are very similar to the previous features?\nThanks.",
      "votes": null
    },
    {
      "id": "2168176",
      "postDate": "03/04/2023 00:50:22",
      "content": "<p>I don't see a reason why the CV should decrease then. But maybe someone more experienced can tell. But for me wouldn't make sense </p>",
      "rawMarkdown": "I don't see a reason why the CV should decrease then. But maybe someone more experienced can tell. But for me wouldn't make sense",
      "votes": null
    },
    {
      "id": "2169361",
      "postDate": "03/05/2023 03:54:09",
      "content": "<p>I think your features  may overfit the train set, or the distribution of these features on the train and test sets is quite different</p>",
      "rawMarkdown": "I think your features  may overfit the train set, or the distribution of these features on the train and test sets is quite different",
      "votes": null
    },
    {
      "id": "2170441",
      "postDate": "03/06/2023 02:15:35",
      "content": "<p>fixing our features helped but we still have that point where nothing seems to help..</p>",
      "rawMarkdown": "fixing our features helped but we still have that point where nothing seems to help..",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2168063,
      "author_name": "simonveitner",
      "author_url": "",
      "post_date": "03/03/2023 20:59:39",
      "content": "<p>I think your features are not good. For me never happens if I use alpha=8</p>",
      "votes": null,
      "replies": [
        {
          "id": 2168069,
          "author_name": "shashwatraman",
          "author_url": "",
          "post_date": "03/03/2023 21:03:49",
          "content": "<p>Yeah maybe. I'll check. Can this also happen if the newly added features are very similar to the previous features?<br>\nThanks.</p>",
          "votes": null,
          "replies": [
            {
              "id": 2168176,
              "author_name": "simonveitner",
              "author_url": "",
              "post_date": "03/04/2023 00:50:22",
              "content": "<p>I don't see a reason why the CV should decrease then. But maybe someone more experienced can tell. But for me wouldn't make sense </p>",
              "votes": null,
              "replies": [
                {
                  "id": 2170441,
                  "author_name": "janmpia",
                  "author_url": "",
                  "post_date": "03/06/2023 02:15:35",
                  "content": "<p>fixing our features helped but we still have that point where nothing seems to help..</p>",
                  "votes": null,
                  "replies": []
                }
              ]
            }
          ]
        }
      ]
    },
    {
      "id": 2169361,
      "author_name": "pysongwang",
      "author_url": "",
      "post_date": "03/05/2023 03:54:09",
      "content": "<p>I think your features  may overfit the train set, or the distribution of these features on the train and test sets is quite different</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2168062": "Hii,\nAfter adding about 80-90 features to the model, a point comes where the score decreases with any new bunch of features added. I am not able to increase the score any further, no matter what I do.\nIs this because the new features are just not helpful, there is too much noise in the previous features or there is some other reason?\nI even tried using alpha parameter of xgboost, but it doesn't help. 😓\nWhat can I do to fix this? \nThanks.",
    "2168063": "I think your features are not good. For me never happens if I use alpha=8",
    "2168069": "Yeah maybe. I'll check. Can this also happen if the newly added features are very similar to the previous features?\nThanks.",
    "2168176": "I don't see a reason why the CV should decrease then. But maybe someone more experienced can tell. But for me wouldn't make sense",
    "2169361": "I think your features  may overfit the train set, or the distribution of these features on the train and test sets is quite different",
    "2170441": "fixing our features helped but we still have that point where nothing seems to help.."
  },
  "source": "meta"
}