{
  "id": 114544,
  "title": "Validation Scheme (general)",
  "url": "/competitions/ashrae-energy-prediction/discussion/114544",
  "author_name": "",
  "post_date": "2019-10-27T07:47:45.506495300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi!</p>\n\n<p>I do struggle in general for Kaggle Competitions to make a good validation scheme. Since it's maybe some of the most crucial aspects I seek some inputs and tips for establishing a robust validation scheme. So, how do you guys go arround that part of the problem? </p>\n\n<p>Do you divide your training data into train and test with holdout, and then using CV for tuning parameters in training set only?</p>\n\n<p>In case, how do you make sure that your holdout test set is similar to LB? </p>\n\n<p>So in general - come up please with some tips, tricks and workflows for establishing a good vld scheme here for Kaggle competitions :) </p>\n\n<p>Thank you all, good competition! </p>",
  "messages": [
    {
      "id": "659172",
      "postDate": "10/27/2019 07:47:45",
      "content": "<p>Hi!</p>\n\n<p>I do struggle in general for Kaggle Competitions to make a good validation scheme. Since it's maybe some of the most crucial aspects I seek some inputs and tips for establishing a robust validation scheme. So, how do you guys go arround that part of the problem? </p>\n\n<p>Do you divide your training data into train and test with holdout, and then using CV for tuning parameters in training set only?</p>\n\n<p>In case, how do you make sure that your holdout test set is similar to LB? </p>\n\n<p>So in general - come up please with some tips, tricks and workflows for establishing a good vld scheme here for Kaggle competitions :) </p>\n\n<p>Thank you all, good competition! </p>",
      "rawMarkdown": "Hi!\n\nI do struggle in general for Kaggle Competitions to make a good validation scheme. Since it's maybe some of the most crucial aspects I seek some inputs and tips for establishing a robust validation scheme. So, how do you guys go arround that part of the problem? \n\nDo you divide your training data into train and test with holdout, and then using CV for tuning parameters in training set only?\n\nIn case, how do you make sure that your holdout test set is similar to LB? \n\nSo in general - come up please with some tips, tricks and workflows for establishing a good vld scheme here for Kaggle competitions :) \n\nThank you all, good competition!",
      "votes": null
    },
    {
      "id": "659233",
      "postDate": "10/27/2019 10:10:08",
      "content": "<p>You may check this thread here for guidelines\n<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931</a></p>",
      "rawMarkdown": "You may check this thread here for guidelines\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931",
      "votes": null
    },
    {
      "id": "659305",
      "postDate": "10/27/2019 12:34:47",
      "content": "<p>Thank. I saw the post also earlier also but was curious to hear some general takes to finding good validation schemes. But thanks for sharing! :) </p>",
      "rawMarkdown": "Thank. I saw the post also earlier also but was curious to hear some general takes to finding good validation schemes. But thanks for sharing! :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 659233,
      "author_name": "imeintanis",
      "author_url": "",
      "post_date": "10/27/2019 10:10:08",
      "content": "<p>You may check this thread here for guidelines\n<a href=\"https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931\">https://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 659305,
          "author_name": "nybyhansen",
          "author_url": "",
          "post_date": "10/27/2019 12:34:47",
          "content": "<p>Thank. I saw the post also earlier also but was curious to hear some general takes to finding good validation schemes. But thanks for sharing! :) </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "659172": "Hi!\n\nI do struggle in general for Kaggle Competitions to make a good validation scheme. Since it's maybe some of the most crucial aspects I seek some inputs and tips for establishing a robust validation scheme. So, how do you guys go arround that part of the problem? \n\nDo you divide your training data into train and test with holdout, and then using CV for tuning parameters in training set only?\n\nIn case, how do you make sure that your holdout test set is similar to LB? \n\nSo in general - come up please with some tips, tricks and workflows for establishing a good vld scheme here for Kaggle competitions :) \n\nThank you all, good competition!",
    "659233": "You may check this thread here for guidelines\nhttps://www.kaggle.com/c/ashrae-energy-prediction/discussion/113644#latest-658931",
    "659305": "Thank. I saw the post also earlier also but was curious to hear some general takes to finding good validation schemes. But thanks for sharing! :)"
  },
  "source": "meta"
}