{
  "id": 50605,
  "title": "Are the distributions of the train and test data different ?",
  "url": "/competitions/data-science-bowl-2018/discussion/50605",
  "author_name": "",
  "post_date": "2018-02-27T00:35:46.596988200Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am currently getting a train and val score of 0.35 and 0.31 but a leaderboard score of 0.15. Is this more likely to be a problem in my code (eg: overfitting) or the train and test distributions being different ? </p>",
  "messages": [
    {
      "id": "288215",
      "postDate": "02/27/2018 00:35:46",
      "content": "<p>I am currently getting a train and val score of 0.35 and 0.31 but a leaderboard score of 0.15. Is this more likely to be a problem in my code (eg: overfitting) or the train and test distributions being different ? </p>",
      "rawMarkdown": "I am currently getting a train and val score of 0.35 and 0.31 but a leaderboard score of 0.15. Is this more likely to be a problem in my code (eg: overfitting) or the train and test distributions being different ?",
      "votes": null
    },
    {
      "id": "288289",
      "postDate": "02/27/2018 05:53:26",
      "content": "<p>Given the magnitude of the difference, it seems that your model is, in fact, suffering from overfitting. </p>\n\n<p>I can't tell how you are calculating val score, but if you aren't doing cross-validation it is possible that the parameters aren't well tuned.  Can you throw more light?</p>",
      "rawMarkdown": "Given the magnitude of the difference, it seems that your model is, in fact, suffering from overfitting. \n\nI can't tell how you are calculating val score, but if you aren't doing cross-validation it is possible that the parameters aren't well tuned.  Can you throw more light?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 288289,
      "author_name": "gangal",
      "author_url": "",
      "post_date": "02/27/2018 05:53:26",
      "content": "<p>Given the magnitude of the difference, it seems that your model is, in fact, suffering from overfitting. </p>\n\n<p>I can't tell how you are calculating val score, but if you aren't doing cross-validation it is possible that the parameters aren't well tuned.  Can you throw more light?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "288215": "I am currently getting a train and val score of 0.35 and 0.31 but a leaderboard score of 0.15. Is this more likely to be a problem in my code (eg: overfitting) or the train and test distributions being different ?",
    "288289": "Given the magnitude of the difference, it seems that your model is, in fact, suffering from overfitting. \n\nI can't tell how you are calculating val score, but if you aren't doing cross-validation it is possible that the parameters aren't well tuned.  Can you throw more light?"
  },
  "source": "meta"
}