{
  "id": 192463,
  "title": "Are you using somehow users which are both in train and test data?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/192463",
  "author_name": "Anaida Shaginian",
  "post_date": "2020-10-21T16:22:49.205000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I found some users which are both in train and test data. I have a plan to split users to smart ones which spend a lot of time on the platform and answer a lot of questions correctly, and not. I want to count it on the train data and then to transform new feature to the test or construct some predictive model for predicting the new feature for missing users. The question is: in which ways do you use the users from train and test both?</p>",
  "messages": [
    {
      "id": 1056364,
      "postDate": "2020-10-21T16:22:49.207Z",
      "content": "<p>I found some users which are both in train and test data. I have a plan to split users to smart ones which spend a lot of time on the platform and answer a lot of questions correctly, and not. I want to count it on the train data and then to transform new feature to the test or construct some predictive model for predicting the new feature for missing users. The question is: in which ways do you use the users from train and test both?</p>",
      "rawMarkdown": "I found some users which are both in train and test data. I have a plan to split users to smart ones which spend a lot of time on the platform and answer a lot of questions correctly, and not. I want to count it on the train data and then to transform new feature to the test or construct some predictive model for predicting the new feature for missing users. The question is: in which ways do you use the users from train and test both?"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1056364": "I found some users which are both in train and test data. I have a plan to split users to smart ones which spend a lot of time on the platform and answer a lot of questions correctly, and not. I want to count it on the train data and then to transform new feature to the test or construct some predictive model for predicting the new feature for missing users. The question is: in which ways do you use the users from train and test both?"
  }
}