{
  "id": 246299,
  "title": "How to predict no dataset for some text ids?",
  "url": "/competitions/coleridgeinitiative-show-us-the-data/discussion/246299",
  "author_name": "",
  "post_date": "2021-06-14T19:43:47.397378200Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>There may be some text ids for which model may not be able to predict any dataset with high confidence. Given the metric penalises false positives much more, we may not want to predict any dataset for those cases. So, is there a way to predict no dataset for some ids? Is it mandatory to predict at least one dataset for each id? What should be PredictionString in cases where we don't want to make any prediction?</p>",
  "messages": [
    {
      "id": "1349468",
      "postDate": "06/14/2021 19:43:47",
      "content": "<p>There may be some text ids for which model may not be able to predict any dataset with high confidence. Given the metric penalises false positives much more, we may not want to predict any dataset for those cases. So, is there a way to predict no dataset for some ids? Is it mandatory to predict at least one dataset for each id? What should be PredictionString in cases where we don't want to make any prediction?</p>",
      "rawMarkdown": "There may be some text ids for which model may not be able to predict any dataset with high confidence. Given the metric penalises false positives much more, we may not want to predict any dataset for those cases. So, is there a way to predict no dataset for some ids? Is it mandatory to predict at least one dataset for each id? What should be PredictionString in cases where we don't want to make any prediction?",
      "votes": null
    },
    {
      "id": "1351855",
      "postDate": "06/16/2021 16:42:19",
      "content": "<p>Just use empty string as your prediction</p>",
      "rawMarkdown": "Just use empty string as your prediction",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1351855,
      "author_name": "nbroad",
      "author_url": "",
      "post_date": "06/16/2021 16:42:19",
      "content": "<p>Just use empty string as your prediction</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1349468": "There may be some text ids for which model may not be able to predict any dataset with high confidence. Given the metric penalises false positives much more, we may not want to predict any dataset for those cases. So, is there a way to predict no dataset for some ids? Is it mandatory to predict at least one dataset for each id? What should be PredictionString in cases where we don't want to make any prediction?",
    "1351855": "Just use empty string as your prediction"
  },
  "source": "meta"
}