{
  "id": 203007,
  "title": "Is LGBM Good for Private Set ?",
  "url": "/competitions/riiid-test-answer-prediction/discussion/203007",
  "author_name": "",
  "post_date": "2020-12-13T07:31:51.887106800Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I see lgbm is doing good with basic aggregate features  giving more than average public LB score .</p>\n<p>But is model trained on these features generalized enough to withstand for private data set ?</p>\n<p>So then i think one should invest time in both the models LGB and  Transformer</p>",
  "messages": [
    {
      "id": "1110903",
      "postDate": "12/13/2020 07:31:51",
      "content": "<p>I see lgbm is doing good with basic aggregate features  giving more than average public LB score .</p>\n<p>But is model trained on these features generalized enough to withstand for private data set ?</p>\n<p>So then i think one should invest time in both the models LGB and  Transformer</p>",
      "rawMarkdown": "I see lgbm is doing good with basic aggregate features  giving more than average public LB score .\n\nBut is model trained on these features generalized enough to withstand for private data set ?\n\nSo then i think one should invest time in both the models LGB and  Transformer",
      "votes": null
    },
    {
      "id": "1111144",
      "postDate": "12/13/2020 13:07:23",
      "content": "<p>in many of my experiences in various competitions, NN outperforms lightgbm in private leaderboard. NN will generalize much better for sure, so if private test set has suprises, then I expect NN to do better :)</p>",
      "rawMarkdown": "in many of my experiences in various competitions, NN outperforms lightgbm in private leaderboard. NN will generalize much better for sure, so if private test set has suprises, then I expect NN to do better :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1111144,
      "author_name": "nikhilmishradev",
      "author_url": "",
      "post_date": "12/13/2020 13:07:23",
      "content": "<p>in many of my experiences in various competitions, NN outperforms lightgbm in private leaderboard. NN will generalize much better for sure, so if private test set has suprises, then I expect NN to do better :)</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1110903": "I see lgbm is doing good with basic aggregate features  giving more than average public LB score .\n\nBut is model trained on these features generalized enough to withstand for private data set ?\n\nSo then i think one should invest time in both the models LGB and  Transformer",
    "1111144": "in many of my experiences in various competitions, NN outperforms lightgbm in private leaderboard. NN will generalize much better for sure, so if private test set has suprises, then I expect NN to do better :)"
  },
  "source": "meta"
}