{
  "id": 72113,
  "title": "Predict Best Model Value",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/72113",
  "author_name": "",
  "post_date": "2018-11-20T13:25:49.320111100Z",
  "votes": 1,
  "comment_count": 6,
  "views": 0,
  "content": "<p>This is a very difficult problem because we are trying to predict the future sales for a short time frame without being able to rely on any of the future data. How much better do you expect the best models to perform compared to just submitting all zeros? If the RMSE for all zeros is 2.000, then I don't think the first place finisher will score better than 1.900. What do you all think?</p>",
  "messages": [
    {
      "id": "424644",
      "postDate": "11/20/2018 13:25:49",
      "content": "<p>This is a very difficult problem because we are trying to predict the future sales for a short time frame without being able to rely on any of the future data. How much better do you expect the best models to perform compared to just submitting all zeros? If the RMSE for all zeros is 2.000, then I don't think the first place finisher will score better than 1.900. What do you all think?</p>",
      "rawMarkdown": "This is a very difficult problem because we are trying to predict the future sales for a short time frame without being able to rely on any of the future data. How much better do you expect the best models to perform compared to just submitting all zeros? If the RMSE for all zeros is 2.000, then I don't think the first place finisher will score better than 1.900. What do you all think?",
      "votes": null
    },
    {
      "id": "424714",
      "postDate": "11/20/2018 15:07:18",
      "content": "<p>I think rmse for all zeros will be around .4 and I agree that the best solution won't be much better. In fact I was thinking about making one submission with all zeros since this problem is so unpredictable.</p>",
      "rawMarkdown": "I think rmse for all zeros will be around .4 and I agree that the best solution won't be much better. In fact I was thinking about making one submission with all zeros since this problem is so unpredictable.",
      "votes": null
    },
    {
      "id": "424731",
      "postDate": "11/20/2018 15:35:01",
      "content": "<p>Given the amount of people who have submitted the public kernel that scores 0.0, I think that a submission of all zeroes will return at least a bronze medal</p>",
      "rawMarkdown": "Given the amount of people who have submitted the public kernel that scores 0.0, I think that a submission of all zeroes will return at least a bronze medal",
      "votes": null
    },
    {
      "id": "424810",
      "postDate": "11/20/2018 17:35:22",
      "content": "<p>Hello Corey,</p>\n\n<p>Could you tell me what makes you think that ? </p>",
      "rawMarkdown": "Hello Corey,\n\nCould you tell me what makes you think that ?",
      "votes": null
    },
    {
      "id": "424844",
      "postDate": "11/20/2018 18:53:57",
      "content": "<p>I think he was being facetious.</p>",
      "rawMarkdown": "I think he was being facetious.",
      "votes": null
    },
    {
      "id": "424845",
      "postDate": "11/20/2018 18:54:21",
      "content": "<p>Hi Y Assous,</p>\n\n<p>It's just a hunch.</p>",
      "rawMarkdown": "Hi Y Assous,\n\nIt's just a hunch.",
      "votes": null
    },
    {
      "id": "427228",
      "postDate": "11/24/2018 22:15:38",
      "content": "<p>I've been having similar thoughts. The problem is further compounded by the lag between the \"test\" data on kaggle and the prediction period. There are 46 days between 10-15-2018 (the last day of the kaggle data set) and 12-01-2018 (the first day of the prediction period). I imagine the best solutions will make heavy use of the demo account to try and close that gap. </p>",
      "rawMarkdown": "I've been having similar thoughts. The problem is further compounded by the lag between the \"test\" data on kaggle and the prediction period. There are 46 days between 10-15-2018 (the last day of the kaggle data set) and 12-01-2018 (the first day of the prediction period). I imagine the best solutions will make heavy use of the demo account to try and close that gap.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 424714,
      "author_name": "jsaguiar",
      "author_url": "",
      "post_date": "11/20/2018 15:07:18",
      "content": "<p>I think rmse for all zeros will be around .4 and I agree that the best solution won't be much better. In fact I was thinking about making one submission with all zeros since this problem is so unpredictable.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 424731,
      "author_name": "returnofsputnik",
      "author_url": "",
      "post_date": "11/20/2018 15:35:01",
      "content": "<p>Given the amount of people who have submitted the public kernel that scores 0.0, I think that a submission of all zeroes will return at least a bronze medal</p>",
      "votes": null,
      "replies": [
        {
          "id": 424810,
          "author_name": "xtelima",
          "author_url": "",
          "post_date": "11/20/2018 17:35:22",
          "content": "<p>Hello Corey,</p>\n\n<p>Could you tell me what makes you think that ? </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 424844,
          "author_name": "afgonczol",
          "author_url": "",
          "post_date": "11/20/2018 18:53:57",
          "content": "<p>I think he was being facetious.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 424845,
          "author_name": "returnofsputnik",
          "author_url": "",
          "post_date": "11/20/2018 18:54:21",
          "content": "<p>Hi Y Assous,</p>\n\n<p>It's just a hunch.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 427228,
      "author_name": "johepaton",
      "author_url": "",
      "post_date": "11/24/2018 22:15:38",
      "content": "<p>I've been having similar thoughts. The problem is further compounded by the lag between the \"test\" data on kaggle and the prediction period. There are 46 days between 10-15-2018 (the last day of the kaggle data set) and 12-01-2018 (the first day of the prediction period). I imagine the best solutions will make heavy use of the demo account to try and close that gap. </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "424644": "This is a very difficult problem because we are trying to predict the future sales for a short time frame without being able to rely on any of the future data. How much better do you expect the best models to perform compared to just submitting all zeros? If the RMSE for all zeros is 2.000, then I don't think the first place finisher will score better than 1.900. What do you all think?",
    "424714": "I think rmse for all zeros will be around .4 and I agree that the best solution won't be much better. In fact I was thinking about making one submission with all zeros since this problem is so unpredictable.",
    "424731": "Given the amount of people who have submitted the public kernel that scores 0.0, I think that a submission of all zeroes will return at least a bronze medal",
    "424810": "Hello Corey,\n\nCould you tell me what makes you think that ?",
    "424844": "I think he was being facetious.",
    "424845": "Hi Y Assous,\n\nIt's just a hunch.",
    "427228": "I've been having similar thoughts. The problem is further compounded by the lag between the \"test\" data on kaggle and the prediction period. There are 46 days between 10-15-2018 (the last day of the kaggle data set) and 12-01-2018 (the first day of the prediction period). I imagine the best solutions will make heavy use of the demo account to try and close that gap."
  },
  "source": "meta"
}