{
  "id": 113959,
  "title": "How can we predict zeros in target variable.",
  "url": "/competitions/ashrae-energy-prediction/discussion/113959",
  "author_name": "JumpeiT",
  "post_date": "2019-10-23T10:11:43.391000",
  "votes": 0,
  "comment_count": 3,
  "views": 0,
  "content": "<p>After I read this article (<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774</a>), \nI consider to regress (or classify) when the target variable becomes zero on some features (for example,  the weather condition, disaster accidents, or power plants' disorders).\nBut it takes a very long journey because identification of building location and data collections are so hard...\nI want to ask you guys the alternative way to predict the zeros events.</p>",
  "messages": [
    {
      "id": 655851,
      "postDate": "2019-10-23T15:18:58.647Z",
      "content": "<p>My model already predicts values less than zero.  I am trying to break it of that bad habit.</p>\n\n<p>Unless your correcting for it, your model probably doing the same right now.</p>",
      "rawMarkdown": "My model already predicts values less than zero.  I am trying to break it of that bad habit.\n\nUnless your correcting for it, your model probably doing the same right now.",
      "votes": 1
    },
    {
      "id": 655684,
      "postDate": "2019-10-23T11:06:51.123Z",
      "content": "<p>You added weather conditions to your model, so the model already has this information and if there is some interaction between failure (target = 0) and whether, the model will recognise that and predict it in future. I don't think that it makes sense to predict target = 0 or not</p>",
      "rawMarkdown": "You added weather conditions to your model, so the model already has this information and if there is some interaction between failure (target = 0) and whether, the model will recognise that and predict it in future. I don't think that it makes sense to predict target = 0 or not"
    },
    {
      "id": 655642,
      "postDate": "2019-10-23T10:11:43.390Z",
      "content": "<p>After I read this article (<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774</a>), \nI consider to regress (or classify) when the target variable becomes zero on some features (for example,  the weather condition, disaster accidents, or power plants' disorders).\nBut it takes a very long journey because identification of building location and data collections are so hard...\nI want to ask you guys the alternative way to predict the zeros events.</p>",
      "rawMarkdown": "After I read this article (https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774), \nI consider to regress (or classify) when the target variable becomes zero on some features (for example,  the weather condition, disaster accidents, or power plants' disorders).\nBut it takes a very long journey because identification of building location and data collections are so hard...\nI want to ask you guys the alternative way to predict the zeros events."
    },
    {
      "id": 655849,
      "postDate": "2019-10-23T15:16:46.563Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 655851,
      "author_name": "PC Jimmmy",
      "author_url": "",
      "post_date": "2019-10-23T15:18:58.647000",
      "content": "<p>My model already predicts values less than zero.  I am trying to break it of that bad habit.</p>\n\n<p>Unless your correcting for it, your model probably doing the same right now.</p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 655684,
      "author_name": "Oleg Knaub",
      "author_url": "",
      "post_date": "2019-10-23T11:06:51.123000",
      "content": "<p>You added weather conditions to your model, so the model already has this information and if there is some interaction between failure (target = 0) and whether, the model will recognise that and predict it in future. I don't think that it makes sense to predict target = 0 or not</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 655849,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-10-23T15:16:46.563000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "655851": "My model already predicts values less than zero.  I am trying to break it of that bad habit.\n\nUnless your correcting for it, your model probably doing the same right now.",
    "655684": "You added weather conditions to your model, so the model already has this information and if there is some interaction between failure (target = 0) and whether, the model will recognise that and predict it in future. I don't think that it makes sense to predict target = 0 or not",
    "655642": "After I read this article (https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113774), \nI consider to regress (or classify) when the target variable becomes zero on some features (for example,  the weather condition, disaster accidents, or power plants' disorders).\nBut it takes a very long journey because identification of building location and data collections are so hard...\nI want to ask you guys the alternative way to predict the zeros events.",
    "655849": ""
  }
}