{
  "id": 195099,
  "title": "How to  handle the unseen user_id in testset",
  "url": "/competitions/riiid-test-answer-prediction/discussion/195099",
  "author_name": "",
  "post_date": "2020-11-03T14:27:14.937234700Z",
  "votes": null,
  "comment_count": 1,
  "views": 0,
  "content": "<p>For me, in local testset, if the use_id data use to generate feature, predict this user other case can easily up to 7.82, but if this featuer not contain this user_id it's just drop to 7.27.<br>\nin the feature importance , user_id feature has a large proportion. So  do you have a best way to deal with user_id you haven't seen before.</p>\n<p>In my opinion, dealing with unknown users_ ID is the key to winning</p>",
  "messages": [
    {
      "id": "1068574",
      "postDate": "11/03/2020 14:27:14",
      "content": "<p>For me, in local testset, if the use_id data use to generate feature, predict this user other case can easily up to 7.82, but if this featuer not contain this user_id it's just drop to 7.27.<br>\nin the feature importance , user_id feature has a large proportion. So  do you have a best way to deal with user_id you haven't seen before.</p>\n<p>In my opinion, dealing with unknown users_ ID is the key to winning</p>",
      "rawMarkdown": "For me, in local testset, if the use_id data use to generate feature, predict this user other case can easily up to 7.82, but if this featuer not contain this user_id it's just drop to 7.27.\nin the feature importance , user_id feature has a large proportion. So  do you have a best way to deal with user_id you haven't seen before.\n\nIn my opinion, dealing with unknown users_ ID is the key to winning",
      "votes": null
    },
    {
      "id": "1069500",
      "postDate": "11/04/2020 14:26:12",
      "content": "<p>I totally agree with you I'm working on how to deal with that!</p>",
      "rawMarkdown": "I totally agree with you I'm working on how to deal with that!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1069500,
      "author_name": "matthieuplante",
      "author_url": "",
      "post_date": "11/04/2020 14:26:12",
      "content": "<p>I totally agree with you I'm working on how to deal with that!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1068574": "For me, in local testset, if the use_id data use to generate feature, predict this user other case can easily up to 7.82, but if this featuer not contain this user_id it's just drop to 7.27.\nin the feature importance , user_id feature has a large proportion. So  do you have a best way to deal with user_id you haven't seen before.\n\nIn my opinion, dealing with unknown users_ ID is the key to winning",
    "1069500": "I totally agree with you I'm working on how to deal with that!"
  },
  "source": "meta"
}