{
  "id": 113240,
  "title": "Splitting models by meter",
  "url": "/competitions/ashrae-energy-prediction/discussion/113240",
  "author_name": "",
  "post_date": "2019-10-18T02:46:44.974953800Z",
  "votes": 6,
  "comment_count": 4,
  "views": 0,
  "content": "<p>The problem gives us 4 different meter types to predict. \nThe distributions of each are fairly different. \nJust wondering if anyone has had better accuracy by splitting the problem into 4 separate regression problems?</p>",
  "messages": [
    {
      "id": "651852",
      "postDate": "10/18/2019 02:46:44",
      "content": "<p>The problem gives us 4 different meter types to predict. \nThe distributions of each are fairly different. \nJust wondering if anyone has had better accuracy by splitting the problem into 4 separate regression problems?</p>",
      "rawMarkdown": "The problem gives us 4 different meter types to predict. \nThe distributions of each are fairly different. \nJust wondering if anyone has had better accuracy by splitting the problem into 4 separate regression problems?",
      "votes": null
    },
    {
      "id": "674865",
      "postDate": "11/17/2019 07:01:59",
      "content": "<p>I tried this before. But there's no advantage with GBM model. Thanks for sharing your idea though.</p>",
      "rawMarkdown": "I tried this before. But there's no advantage with GBM model. Thanks for sharing your idea though.",
      "votes": null
    },
    {
      "id": "675160",
      "postDate": "11/17/2019 17:18:42",
      "content": "<p>Currently working with splitting using lgbm.  Once I get a decent model I plan to build separate feature data sets for each meter.  Aiden has seen no advantage with gbm model - I have seen advantage (molecules competition) when each part of the problem has it's own feature set.   </p>\n\n<p>It does make for a more confusing world however as I struggle keeping the features sets straight.</p>\n\n<p>I working with fork of this one\n<a href=\"https://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type\">https://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type</a></p>",
      "rawMarkdown": "Currently working with splitting using lgbm.  Once I get a decent model I plan to build separate feature data sets for each meter.  Aiden has seen no advantage with gbm model - I have seen advantage (molecules competition) when each part of the problem has it's own feature set.   \n\nIt does make for a more confusing world however as I struggle keeping the features sets straight.\n\nI working with fork of this one\nhttps://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type",
      "votes": null
    },
    {
      "id": "675315",
      "postDate": "11/17/2019 22:41:29",
      "content": "<p>Thanks for sharing !</p>",
      "rawMarkdown": "Thanks for sharing !",
      "votes": null
    },
    {
      "id": "697792",
      "postDate": "12/18/2019 12:14:58",
      "content": "<p>I think the point is consider not only the meter but also the number of meters present at the building. If there are &gt;1 energy sources available,  consumption should be distributed among sources </p>",
      "rawMarkdown": "I think the point is consider not only the meter but also the number of meters present at the building. If there are &gt;1 energy sources available,  consumption should be distributed among sources",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 674865,
      "author_name": "oseongchoi",
      "author_url": "",
      "post_date": "11/17/2019 07:01:59",
      "content": "<p>I tried this before. But there's no advantage with GBM model. Thanks for sharing your idea though.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 675160,
      "author_name": "pcjimmmy",
      "author_url": "",
      "post_date": "11/17/2019 17:18:42",
      "content": "<p>Currently working with splitting using lgbm.  Once I get a decent model I plan to build separate feature data sets for each meter.  Aiden has seen no advantage with gbm model - I have seen advantage (molecules competition) when each part of the problem has it's own feature set.   </p>\n\n<p>It does make for a more confusing world however as I struggle keeping the features sets straight.</p>\n\n<p>I working with fork of this one\n<a href=\"https://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type\">https://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 675315,
          "author_name": "oseongchoi",
          "author_url": "",
          "post_date": "11/17/2019 22:41:29",
          "content": "<p>Thanks for sharing !</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 697792,
      "author_name": "lauramorenosarrion",
      "author_url": "",
      "post_date": "12/18/2019 12:14:58",
      "content": "<p>I think the point is consider not only the meter but also the number of meters present at the building. If there are &gt;1 energy sources available,  consumption should be distributed among sources </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "651852": "The problem gives us 4 different meter types to predict. \nThe distributions of each are fairly different. \nJust wondering if anyone has had better accuracy by splitting the problem into 4 separate regression problems?",
    "674865": "I tried this before. But there's no advantage with GBM model. Thanks for sharing your idea though.",
    "675160": "Currently working with splitting using lgbm.  Once I get a decent model I plan to build separate feature data sets for each meter.  Aiden has seen no advantage with gbm model - I have seen advantage (molecules competition) when each part of the problem has it's own feature set.   \n\nIt does make for a more confusing world however as I struggle keeping the features sets straight.\n\nI working with fork of this one\nhttps://www.kaggle.com/corochann/ashrae-training-lgbm-by-meter-type",
    "675315": "Thanks for sharing !",
    "697792": "I think the point is consider not only the meter but also the number of meters present at the building. If there are &gt;1 energy sources available,  consumption should be distributed among sources"
  },
  "source": "meta"
}