{
  "id": 335423,
  "title": "How to make your Model More robust ?",
  "url": "/competitions/amex-default-prediction/discussion/335423",
  "author_name": "",
  "post_date": "2022-07-06T04:36:27.058307500Z",
  "votes": 13,
  "comment_count": 2,
  "views": 0,
  "content": "<p>As we can see in public leaderboard, there are lot of competitors with similar scores Rank 60 to rank 437 around 377 participants have same score on Public leaderboard as 0.797. So a minor change in your score on private leaderboard can make a huge difference in your final Rank in this Competition.  <br>\nSo what all we can do to make our solution more robust, so that we don't slide down in private leader board.<br>\nSome of the top things in my mind are following:</p>\n<ol>\n<li>Using Multiple model ensemble to make your model more robust.</li>\n<li>Using K fold to make your model more robust.</li>\n<li>Build multiple models with different seeds and ensemble them.</li>\n<li>Use adversarial Validation to make your model more robust. Link to get details on Adversarial validation- <a href=\"https://www.kaggle.com/competitions/amex-default-prediction/discussion/335398\" target=\"_blank\">link</a></li>\n</ol>\n<p>Please comment below and let me know if any thing I have missed . And don't forget to upvote this discussion if you feel its helpful.</p>",
  "messages": [
    {
      "id": "1845117",
      "postDate": "07/06/2022 04:36:27",
      "content": "<p>As we can see in public leaderboard, there are lot of competitors with similar scores Rank 60 to rank 437 around 377 participants have same score on Public leaderboard as 0.797. So a minor change in your score on private leaderboard can make a huge difference in your final Rank in this Competition.  <br>\nSo what all we can do to make our solution more robust, so that we don't slide down in private leader board.<br>\nSome of the top things in my mind are following:</p>\n<ol>\n<li>Using Multiple model ensemble to make your model more robust.</li>\n<li>Using K fold to make your model more robust.</li>\n<li>Build multiple models with different seeds and ensemble them.</li>\n<li>Use adversarial Validation to make your model more robust. Link to get details on Adversarial validation- <a href=\"https://www.kaggle.com/competitions/amex-default-prediction/discussion/335398\" target=\"_blank\">link</a></li>\n</ol>\n<p>Please comment below and let me know if any thing I have missed . And don't forget to upvote this discussion if you feel its helpful.</p>",
      "rawMarkdown": "As we can see in public leaderboard, there are lot of competitors with similar scores Rank 60 to rank 437 around 377 participants have same score on Public leaderboard as 0.797. So a minor change in your score on private leaderboard can make a huge difference in your final Rank in this Competition.  \nSo what all we can do to make our solution more robust, so that we don't slide down in private leader board.\nSome of the top things in my mind are following:\n1. Using Multiple model ensemble to make your model more robust.\n2. Using K fold to make your model more robust.\n3. Build multiple models with different seeds and ensemble them.\n4. Use adversarial Validation to make your model more robust. Link to get details on Adversarial validation- [link](https://www.kaggle.com/competitions/amex-default-prediction/discussion/335398)\n\nPlease comment below and let me know if any thing I have missed . And don't forget to upvote this discussion if you feel its helpful.",
      "votes": null
    },
    {
      "id": "1846345",
      "postDate": "07/07/2022 03:13:33",
      "content": "<p>Start with error analysis: Check where exactly your models are wrong. (and why?)</p>",
      "rawMarkdown": "Start with error analysis: Check where exactly your models are wrong. (and why?)",
      "votes": null
    },
    {
      "id": "1846496",
      "postDate": "07/07/2022 05:42:19",
      "content": "<p>any suggestion/example on error analysis?</p>",
      "rawMarkdown": "any suggestion/example on error analysis?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1846345,
      "author_name": "thedevastator",
      "author_url": "",
      "post_date": "07/07/2022 03:13:33",
      "content": "<p>Start with error analysis: Check where exactly your models are wrong. (and why?)</p>",
      "votes": null,
      "replies": [
        {
          "id": 1846496,
          "author_name": "smnomaan",
          "author_url": "",
          "post_date": "07/07/2022 05:42:19",
          "content": "<p>any suggestion/example on error analysis?</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "1845117": "As we can see in public leaderboard, there are lot of competitors with similar scores Rank 60 to rank 437 around 377 participants have same score on Public leaderboard as 0.797. So a minor change in your score on private leaderboard can make a huge difference in your final Rank in this Competition.  \nSo what all we can do to make our solution more robust, so that we don't slide down in private leader board.\nSome of the top things in my mind are following:\n1. Using Multiple model ensemble to make your model more robust.\n2. Using K fold to make your model more robust.\n3. Build multiple models with different seeds and ensemble them.\n4. Use adversarial Validation to make your model more robust. Link to get details on Adversarial validation- [link](https://www.kaggle.com/competitions/amex-default-prediction/discussion/335398)\n\nPlease comment below and let me know if any thing I have missed . And don't forget to upvote this discussion if you feel its helpful.",
    "1846345": "Start with error analysis: Check where exactly your models are wrong. (and why?)",
    "1846496": "any suggestion/example on error analysis?"
  },
  "source": "meta"
}