{
  "id": 122366,
  "title": "What is your luckiest multiplier?",
  "url": "/competitions/ashrae-energy-prediction/discussion/122366",
  "author_name": "Nika Zhidkonozhkina",
  "post_date": "2019-12-19T16:52:30.462000",
  "votes": 4,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Probably a lot of participants know that it is better to multiply predictions by a coefficient less than one.</p>",
  "messages": [
    {
      "id": 698746,
      "postDate": "2019-12-19T16:52:30.463Z",
      "content": "<p>Probably a lot of participants know that it is better to multiply predictions by a coefficient less than one.</p>",
      "rawMarkdown": "Probably a lot of participants know that it is better to multiply predictions by a coefficient less than one.",
      "votes": 4
    },
    {
      "id": 699424,
      "postDate": "2019-12-20T12:47:45.930Z",
      "content": "<p>I didn't discover this coefficient until this post mention it. :(</p>",
      "rawMarkdown": "I didn't discover this coefficient until this post mention it. :("
    },
    {
      "id": 699163,
      "postDate": "2019-12-20T05:41:21.747Z",
      "content": "<p>0.9 : I discovered this effect thanks to a coding error :-D \nFrom the leak data you can see that the average energy consumption went down over the years, although not for all meter types/buildings individually</p>",
      "rawMarkdown": "0.9 : I discovered this effect thanks to a coding error :-D \nFrom the leak data you can see that the average energy consumption went down over the years, although not for all meter types/buildings individually"
    },
    {
      "id": 699085,
      "postDate": "2019-12-20T03:23:23.723Z",
      "content": "<p>I found this trick when averaging models. The multiply coefficient can between 0.88 to 0.90.</p>",
      "rawMarkdown": "I found this trick when averaging models. The multiply coefficient can between 0.88 to 0.90."
    },
    {
      "id": 699003,
      "postDate": "2019-12-20T01:02:17.527Z",
      "content": "<p>0.9</p>",
      "rawMarkdown": "0.9"
    },
    {
      "id": 698753,
      "postDate": "2019-12-19T16:56:59.657Z",
      "rawMarkdown": "",
      "isDeleted": true,
      "replies": [
        {
          "id": 698772,
          "postDate": "2019-12-19T17:17:51.140Z",
          "content": "<p>This is all due to the presence of noise in the data. Therefore, we have optimal predicted values between the average of \"noisy\" training meter readings and zero.</p>",
          "rawMarkdown": "This is all due to the presence of noise in the data. Therefore, we have optimal predicted values between the average of \"noisy\" training meter readings and zero.",
          "votes": 1
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 699424,
      "author_name": "Dean",
      "author_url": "",
      "post_date": "2019-12-20T12:47:45.930000",
      "content": "<p>I didn't discover this coefficient until this post mention it. :(</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 699163,
      "author_name": "Rahul",
      "author_url": "",
      "post_date": "2019-12-20T05:41:21.747000",
      "content": "<p>0.9 : I discovered this effect thanks to a coding error :-D \nFrom the leak data you can see that the average energy consumption went down over the years, although not for all meter types/buildings individually</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 699085,
      "author_name": "Yuan Liu",
      "author_url": "",
      "post_date": "2019-12-20T03:23:23.723000",
      "content": "<p>I found this trick when averaging models. The multiply coefficient can between 0.88 to 0.90.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 699003,
      "author_name": "Manraj Singh",
      "author_url": "",
      "post_date": "2019-12-20T01:02:17.527000",
      "content": "<p>0.9</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 698753,
      "author_name": "",
      "author_url": "",
      "post_date": "2019-12-19T16:56:59.657000",
      "content": "",
      "votes": 0,
      "replies": [
        {
          "id": 698772,
          "author_name": "Nika Zhidkonozhkina",
          "author_url": "",
          "post_date": "2019-12-19T17:17:51.140000",
          "content": "<p>This is all due to the presence of noise in the data. Therefore, we have optimal predicted values between the average of \"noisy\" training meter readings and zero.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "698746": "Probably a lot of participants know that it is better to multiply predictions by a coefficient less than one.",
    "699424": "I didn't discover this coefficient until this post mention it. :(",
    "699163": "0.9 : I discovered this effect thanks to a coding error :-D \nFrom the leak data you can see that the average energy consumption went down over the years, although not for all meter types/buildings individually",
    "699085": "I found this trick when averaging models. The multiply coefficient can between 0.88 to 0.90.",
    "699003": "0.9",
    "698753": ""
  }
}