{
  "id": 70618,
  "title": "Terrible in test dataset!!!",
  "url": "/competitions/ga-customer-revenue-prediction/discussion/70618",
  "author_name": "",
  "post_date": "2018-11-05T22:34:16.698726500Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I could get RMSE of train and validation dataset near 1.20.\nBut why the result of the test dataset is so terrible. The score is almost 1.7 !!!</p>",
  "messages": [
    {
      "id": "415910",
      "postDate": "11/05/2018 22:34:16",
      "content": "<p>I could get RMSE of train and validation dataset near 1.20.\nBut why the result of the test dataset is so terrible. The score is almost 1.7 !!!</p>",
      "rawMarkdown": "I could get RMSE of train and validation dataset near 1.20.\nBut why the result of the test dataset is so terrible. The score is almost 1.7 !!!",
      "votes": null
    },
    {
      "id": "416098",
      "postDate": "11/06/2018 07:30:35",
      "content": "<p>Overfitting may be a reason. Check following link <a href=\"http://statisticsbyjim.com/regression/overfitting-regression-models/\">http://statisticsbyjim.com/regression/overfitting-regression-models/</a></p>",
      "rawMarkdown": "Overfitting may be a reason. Check following link http://statisticsbyjim.com/regression/overfitting-regression-models/",
      "votes": null
    },
    {
      "id": "416255",
      "postDate": "11/06/2018 13:44:11",
      "content": "<p>Hello Rachel,</p>\n\n<p>Your model may be overfitting.. but if you look at other kernels in this competition, there is often a big gap between train and validation set..</p>",
      "rawMarkdown": "Hello Rachel,\n\nYour model may be overfitting.. but if you look at other kernels in this competition, there is often a big gap between train and validation set..",
      "votes": null
    },
    {
      "id": "416541",
      "postDate": "11/06/2018 21:05:30",
      "content": "<p>At first, I also think the model is overfitting. But I try to use a more complex model. This time the RMSE of train and validation is higher, 1.50 and 1.6. However, I get a better score of the submission, 1.48. \nSo weird!</p>",
      "rawMarkdown": "At first, I also think the model is overfitting. But I try to use a more complex model. This time the RMSE of train and validation is higher, 1.50 and 1.6. However, I get a better score of the submission, 1.48. \nSo weird!",
      "votes": null
    },
    {
      "id": "416542",
      "postDate": "11/06/2018 21:06:41",
      "content": "<p>But I try to use a more complex model. The score is better. So weird!</p>",
      "rawMarkdown": "But I try to use a more complex model. The score is better. So weird!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 416098,
      "author_name": "rahullalu",
      "author_url": "",
      "post_date": "11/06/2018 07:30:35",
      "content": "<p>Overfitting may be a reason. Check following link <a href=\"http://statisticsbyjim.com/regression/overfitting-regression-models/\">http://statisticsbyjim.com/regression/overfitting-regression-models/</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 416541,
          "author_name": "rachel115",
          "author_url": "",
          "post_date": "11/06/2018 21:05:30",
          "content": "<p>At first, I also think the model is overfitting. But I try to use a more complex model. This time the RMSE of train and validation is higher, 1.50 and 1.6. However, I get a better score of the submission, 1.48. \nSo weird!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 416255,
      "author_name": "xtelima",
      "author_url": "",
      "post_date": "11/06/2018 13:44:11",
      "content": "<p>Hello Rachel,</p>\n\n<p>Your model may be overfitting.. but if you look at other kernels in this competition, there is often a big gap between train and validation set..</p>",
      "votes": null,
      "replies": [
        {
          "id": 416542,
          "author_name": "rachel115",
          "author_url": "",
          "post_date": "11/06/2018 21:06:41",
          "content": "<p>But I try to use a more complex model. The score is better. So weird!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "415910": "I could get RMSE of train and validation dataset near 1.20.\nBut why the result of the test dataset is so terrible. The score is almost 1.7 !!!",
    "416098": "Overfitting may be a reason. Check following link http://statisticsbyjim.com/regression/overfitting-regression-models/",
    "416255": "Hello Rachel,\n\nYour model may be overfitting.. but if you look at other kernels in this competition, there is often a big gap between train and validation set..",
    "416541": "At first, I also think the model is overfitting. But I try to use a more complex model. This time the RMSE of train and validation is higher, 1.50 and 1.6. However, I get a better score of the submission, 1.48. \nSo weird!",
    "416542": "But I try to use a more complex model. The score is better. So weird!"
  },
  "source": "meta"
}