{
  "id": 294797,
  "title": "Using Boosting to Improve Base Models",
  "url": "/competitions/cs5785-fall-2021-final/discussion/294797",
  "author_name": "",
  "post_date": "2021-12-12T21:46:33.902571300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I'm exploring using boosting with AdaBoost to improve the performance of the base ridge regression model. Has anyone attempted to do something similar or ran into the issue where AdaBoostRegressor requires y to be a 1d array, even when the base ridge model doesn't? Would appreciate any advice on this.</p>",
  "messages": [
    {
      "id": "1615893",
      "postDate": "12/12/2021 21:46:33",
      "content": "<p>I'm exploring using boosting with AdaBoost to improve the performance of the base ridge regression model. Has anyone attempted to do something similar or ran into the issue where AdaBoostRegressor requires y to be a 1d array, even when the base ridge model doesn't? Would appreciate any advice on this.</p>",
      "rawMarkdown": "I'm exploring using boosting with AdaBoost to improve the performance of the base ridge regression model. Has anyone attempted to do something similar or ran into the issue where AdaBoostRegressor requires y to be a 1d array, even when the base ridge model doesn't? Would appreciate any advice on this.",
      "votes": null
    },
    {
      "id": "1615921",
      "postDate": "12/12/2021 23:11:34",
      "content": "<p>Try this:</p>\n<p>from sklearn.multioutput import MultiOutputRegressor</p>\n<p>Then wrap the boosted model you're using, so it looks like:</p>\n<p>boosted_model = MultiOutputRegressor(boosted_model)</p>\n<p>Depending on what packages you're using you might need to play around with the syntax a bit, but the general idea should still work. </p>",
      "rawMarkdown": "Try this:\n\nfrom sklearn.multioutput import MultiOutputRegressor\n\nThen wrap the boosted model you're using, so it looks like:\n\nboosted_model = MultiOutputRegressor(boosted_model)\n\nDepending on what packages you're using you might need to play around with the syntax a bit, but the general idea should still work.",
      "votes": null
    },
    {
      "id": "1617430",
      "postDate": "12/14/2021 04:43:42",
      "content": "<p>Thanks! We found that using boosting did improve performance, although not by much (1%)</p>",
      "rawMarkdown": "Thanks! We found that using boosting did improve performance, although not by much (1%)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1615921,
      "author_name": "epickens",
      "author_url": "",
      "post_date": "12/12/2021 23:11:34",
      "content": "<p>Try this:</p>\n<p>from sklearn.multioutput import MultiOutputRegressor</p>\n<p>Then wrap the boosted model you're using, so it looks like:</p>\n<p>boosted_model = MultiOutputRegressor(boosted_model)</p>\n<p>Depending on what packages you're using you might need to play around with the syntax a bit, but the general idea should still work. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1617430,
      "author_name": "hoangdang62",
      "author_url": "",
      "post_date": "12/14/2021 04:43:42",
      "content": "<p>Thanks! We found that using boosting did improve performance, although not by much (1%)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1615893": "I'm exploring using boosting with AdaBoost to improve the performance of the base ridge regression model. Has anyone attempted to do something similar or ran into the issue where AdaBoostRegressor requires y to be a 1d array, even when the base ridge model doesn't? Would appreciate any advice on this.",
    "1615921": "Try this:\n\nfrom sklearn.multioutput import MultiOutputRegressor\n\nThen wrap the boosted model you're using, so it looks like:\n\nboosted_model = MultiOutputRegressor(boosted_model)\n\nDepending on what packages you're using you might need to play around with the syntax a bit, but the general idea should still work.",
    "1617430": "Thanks! We found that using boosting did improve performance, although not by much (1%)"
  },
  "source": "meta"
}