{
  "id": 177845,
  "title": "Stratified GroupKFold usage",
  "url": "/competitions/osic-pulmonary-fibrosis-progression/discussion/177845",
  "author_name": "",
  "post_date": "2020-08-27T14:34:16.327915Z",
  "votes": null,
  "comment_count": 4,
  "views": 0,
  "content": "<p>I wonder is anyone managed to use Stratified GroupKFold to do NN training on the tabular data? </p>",
  "messages": [
    {
      "id": "987808",
      "postDate": "08/27/2020 14:34:16",
      "content": "<p>I wonder is anyone managed to use Stratified GroupKFold to do NN training on the tabular data? </p>",
      "rawMarkdown": "I wonder is anyone managed to use Stratified GroupKFold to do NN training on the tabular data?",
      "votes": null
    },
    {
      "id": "987929",
      "postDate": "08/27/2020 16:15:21",
      "content": "<p>Well, I think it is important that you don't use a patient in your trainset and use some other data from that same patient in your validationset. Is that what you mean? Because then it is doable.</p>",
      "rawMarkdown": "Well, I think it is important that you don't use a patient in your trainset and use some other data from that same patient in your validationset. Is that what you mean? Because then it is doable.",
      "votes": null
    },
    {
      "id": "987981",
      "postDate": "08/27/2020 17:09:06",
      "content": "<p>I'm using a cv that you may refer as Stratified GroupKFold. Patient groups are stratified by Sex and SmokingStatus.</p>",
      "rawMarkdown": "I'm using a cv that you may refer as Stratified GroupKFold. Patient groups are stratified by Sex and SmokingStatus.",
      "votes": null
    },
    {
      "id": "988041",
      "postDate": "08/27/2020 18:10:24",
      "content": "<p>Yes. But it didn't work the magics for me</p>",
      "rawMarkdown": "Yes. But it didn't work the magics for me",
      "votes": null
    },
    {
      "id": "989257",
      "postDate": "08/28/2020 17:01:03",
      "content": "<p>I'll try that, thanks.</p>",
      "rawMarkdown": "I'll try that, thanks.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 987929,
      "author_name": "lukereijnen",
      "author_url": "",
      "post_date": "08/27/2020 16:15:21",
      "content": "<p>Well, I think it is important that you don't use a patient in your trainset and use some other data from that same patient in your validationset. Is that what you mean? Because then it is doable.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 987981,
      "author_name": "gunesevitan",
      "author_url": "",
      "post_date": "08/27/2020 17:09:06",
      "content": "<p>I'm using a cv that you may refer as Stratified GroupKFold. Patient groups are stratified by Sex and SmokingStatus.</p>",
      "votes": null,
      "replies": [
        {
          "id": 989257,
          "author_name": "atamazian",
          "author_url": "",
          "post_date": "08/28/2020 17:01:03",
          "content": "<p>I'll try that, thanks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 988041,
      "author_name": "ulrich07",
      "author_url": "",
      "post_date": "08/27/2020 18:10:24",
      "content": "<p>Yes. But it didn't work the magics for me</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "987808": "I wonder is anyone managed to use Stratified GroupKFold to do NN training on the tabular data?",
    "987929": "Well, I think it is important that you don't use a patient in your trainset and use some other data from that same patient in your validationset. Is that what you mean? Because then it is doable.",
    "987981": "I'm using a cv that you may refer as Stratified GroupKFold. Patient groups are stratified by Sex and SmokingStatus.",
    "988041": "Yes. But it didn't work the magics for me",
    "989257": "I'll try that, thanks."
  },
  "source": "meta"
}