{
  "id": 192616,
  "title": "can we use model aggregation and cross-validation",
  "url": "/competitions/riiid-test-answer-prediction/discussion/192616",
  "author_name": "",
  "post_date": "2020-10-22T11:45:42.061893300Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Is it not possible to use model aggregation and cross-validation in this competition？because the test set can only be called once</p>",
  "messages": [
    {
      "id": "1057124",
      "postDate": "10/22/2020 11:45:42",
      "content": "<p>Is it not possible to use model aggregation and cross-validation in this competition？because the test set can only be called once</p>",
      "rawMarkdown": "Is it not possible to use model aggregation and cross-validation in this competition？because the test set can only be called once",
      "votes": null
    },
    {
      "id": "1057227",
      "postDate": "10/22/2020 13:45:16",
      "content": "<p>Only the data used to submit to the competition can be called once.<br>\nFor cross-validation I would have thought you would be using the training dataset.<br>\nHere is an example from <a href=\"https://www.kaggle.com/mrutyunjaybiswal\" target=\"_blank\">@mrutyunjaybiswal</a> <br>\n<a href=\"https://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv\" target=\"_blank\">https://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv</a><br>\nI hope this helps</p>",
      "rawMarkdown": "Only the data used to submit to the competition can be called once.\nFor cross-validation I would have thought you would be using the training dataset.\nHere is an example from @mrutyunjaybiswal \nhttps://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv\nI hope this helps",
      "votes": null
    },
    {
      "id": "1057238",
      "postDate": "10/22/2020 13:55:40",
      "content": "<p><a href=\"https://www.kaggle.com/ghostskipper\" target=\"_blank\">@ghostskipper</a> Thanks for mentioning buddy. My thought of cross validation was basically due to 2:1 imbalanced Data (if we choose to call it imbalanced) and 80:20 split of train-test set. I have used Stratified KFold as of now, wonder how will it perform for a very basic model with more splits and other CV methods.</p>",
      "rawMarkdown": "ghostskipper Thanks for mentioning buddy. My thought of cross validation was basically due to 2:1 imbalanced Data (if we choose to call it imbalanced) and 80:20 split of train-test set. I have used Stratified KFold as of now, wonder how will it perform for a very basic model with more splits and other CV methods.",
      "votes": null
    },
    {
      "id": "1058000",
      "postDate": "10/23/2020 08:37:01",
      "content": "<p>thanks!!!!!</p>",
      "rawMarkdown": "thanks!!!!!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1057227,
      "author_name": "ghostskipper",
      "author_url": "",
      "post_date": "10/22/2020 13:45:16",
      "content": "<p>Only the data used to submit to the competition can be called once.<br>\nFor cross-validation I would have thought you would be using the training dataset.<br>\nHere is an example from <a href=\"https://www.kaggle.com/mrutyunjaybiswal\" target=\"_blank\">@mrutyunjaybiswal</a> <br>\n<a href=\"https://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv\" target=\"_blank\">https://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv</a><br>\nI hope this helps</p>",
      "votes": null,
      "replies": [
        {
          "id": 1057238,
          "author_name": "mrutyunjaybiswal",
          "author_url": "",
          "post_date": "10/22/2020 13:55:40",
          "content": "<p><a href=\"https://www.kaggle.com/ghostskipper\" target=\"_blank\">@ghostskipper</a> Thanks for mentioning buddy. My thought of cross validation was basically due to 2:1 imbalanced Data (if we choose to call it imbalanced) and 80:20 split of train-test set. I have used Stratified KFold as of now, wonder how will it perform for a very basic model with more splits and other CV methods.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1058000,
          "author_name": "xutianyu123",
          "author_url": "",
          "post_date": "10/23/2020 08:37:01",
          "content": "<p>thanks!!!!!</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1057124": "Is it not possible to use model aggregation and cross-validation in this competition？because the test set can only be called once",
    "1057227": "Only the data used to submit to the competition can be called once.\nFor cross-validation I would have thought you would be using the training dataset.\nHere is an example from @mrutyunjaybiswal \nhttps://www.kaggle.com/mrutyunjaybiswal/riiid-neural-nets-starter-baseline-in-5f-cv\nI hope this helps",
    "1057238": "ghostskipper Thanks for mentioning buddy. My thought of cross validation was basically due to 2:1 imbalanced Data (if we choose to call it imbalanced) and 80:20 split of train-test set. I have used Stratified KFold as of now, wonder how will it perform for a very basic model with more splits and other CV methods.",
    "1058000": "thanks!!!!!"
  },
  "source": "meta"
}