{
  "id": 501552,
  "title": "The strategy for dividing the training set and validation set?",
  "url": "/competitions/leap-atmospheric-physics-ai-climsim/discussion/501552",
  "author_name": "Zhuoqun Li",
  "post_date": "2024-05-09T16:56:40.592000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I found that under the same algorithm, different strategies for dividing the training set and validation set can result in drastically different test scores. This may be due to significant differences in the distribution of certain features between the training and test sets. Does anyone have any good methods for addressing this?</p>",
  "messages": [
    {
      "id": 2803868,
      "postDate": "2024-05-09T16:56:40.593Z",
      "content": "<p>I found that under the same algorithm, different strategies for dividing the training set and validation set can result in drastically different test scores. This may be due to significant differences in the distribution of certain features between the training and test sets. Does anyone have any good methods for addressing this?</p>",
      "rawMarkdown": "I found that under the same algorithm, different strategies for dividing the training set and validation set can result in drastically different test scores. This may be due to significant differences in the distribution of certain features between the training and test sets. Does anyone have any good methods for addressing this?",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2803868": "I found that under the same algorithm, different strategies for dividing the training set and validation set can result in drastically different test scores. This may be due to significant differences in the distribution of certain features between the training and test sets. Does anyone have any good methods for addressing this?"
  }
}