{
  "id": 202063,
  "title": "is this user base competition or content base?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/202063",
  "author_name": "Vishal Vanpariya",
  "post_date": "2020-12-08T05:53:23.722000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>Hi Kagglers,</p>\n<p>Hope all doing great,</p>\n<p>Here in the training dataset, we can see that we have different Userid and Contentid.<br>\nas I read somewhere in a discussion section there will be new users in a private dataset. that means we don't need to use any feature engineering through Userid, we only use Contentid for feature engineering. but one thing is that if we use the content only then it won't be that much good model because the user's learning capacity will differ from other users.</p>\n<p>Machine learning is not about a particular user but here it looks like we can not use a generalized approach for all users.<br>\nMaybe Userid will have a big role in prediction if this competition is based on both user and content. </p>\n<p>Thanks<br>\nVishal</p>",
  "messages": [
    {
      "id": 1105711,
      "postDate": "2020-12-08T05:53:23.723Z",
      "content": "<p>Hi Kagglers,</p>\n<p>Hope all doing great,</p>\n<p>Here in the training dataset, we can see that we have different Userid and Contentid.<br>\nas I read somewhere in a discussion section there will be new users in a private dataset. that means we don't need to use any feature engineering through Userid, we only use Contentid for feature engineering. but one thing is that if we use the content only then it won't be that much good model because the user's learning capacity will differ from other users.</p>\n<p>Machine learning is not about a particular user but here it looks like we can not use a generalized approach for all users.<br>\nMaybe Userid will have a big role in prediction if this competition is based on both user and content. </p>\n<p>Thanks<br>\nVishal</p>",
      "rawMarkdown": "Hi Kagglers,\n\nHope all doing great,\n\nHere in the training dataset, we can see that we have different Userid and Contentid.\nas I read somewhere in a discussion section there will be new users in a private dataset. that means we don't need to use any feature engineering through Userid, we only use Contentid for feature engineering. but one thing is that if we use the content only then it won't be that much good model because the user's learning capacity will differ from other users.\n\nMachine learning is not about a particular user but here it looks like we can not use a generalized approach for all users.\nMaybe Userid will have a big role in prediction if this competition is based on both user and content. \n\nThanks\nVishal"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "1105711": "Hi Kagglers,\n\nHope all doing great,\n\nHere in the training dataset, we can see that we have different Userid and Contentid.\nas I read somewhere in a discussion section there will be new users in a private dataset. that means we don't need to use any feature engineering through Userid, we only use Contentid for feature engineering. but one thing is that if we use the content only then it won't be that much good model because the user's learning capacity will differ from other users.\n\nMachine learning is not about a particular user but here it looks like we can not use a generalized approach for all users.\nMaybe Userid will have a big role in prediction if this competition is based on both user and content. \n\nThanks\nVishal"
  }
}