{
  "id": 79369,
  "title": "Can I use Boruta with only new features ",
  "url": "/competitions/elo-merchant-category-recommendation/discussion/79369",
  "author_name": "tatsu",
  "post_date": "2019-02-03T05:10:40.468000",
  "votes": 2,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi, I am new to kaggle and this is the first competition I have been devoting to. </p>\n\n<p>Since it seems feature selection is so important in this competition, I've implemented boruta with lightGBM.  But it takes so much time to finish with my local computer. So I'm considering the ways to reduce computation time. It would be nice if you tell me whether it is able to select accurately even if using boruta not with all features but with only new features(and target values) .</p>",
  "messages": [
    {
      "id": 465439,
      "postDate": "2019-02-03T05:10:40.470Z",
      "content": "<p>Hi, I am new to kaggle and this is the first competition I have been devoting to. </p>\n\n<p>Since it seems feature selection is so important in this competition, I've implemented boruta with lightGBM.  But it takes so much time to finish with my local computer. So I'm considering the ways to reduce computation time. It would be nice if you tell me whether it is able to select accurately even if using boruta not with all features but with only new features(and target values) .</p>",
      "rawMarkdown": "Hi, I am new to kaggle and this is the first competition I have been devoting to. \n\nSince it seems feature selection is so important in this competition, I've implemented boruta with lightGBM.  But it takes so much time to finish with my local computer. So I'm considering the ways to reduce computation time. It would be nice if you tell me whether it is able to select accurately even if using boruta not with all features but with only new features(and target values) .",
      "votes": 2
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "465439": "Hi, I am new to kaggle and this is the first competition I have been devoting to. \n\nSince it seems feature selection is so important in this competition, I've implemented boruta with lightGBM.  But it takes so much time to finish with my local computer. So I'm considering the ways to reduce computation time. It would be nice if you tell me whether it is able to select accurately even if using boruta not with all features but with only new features(and target values) ."
  }
}