{
  "id": 247514,
  "title": "What does it take to reach 1.000 AUC?",
  "url": "/competitions/seti-breakthrough-listen/discussion/247514",
  "author_name": "FelipeKitamura, MD, PhD",
  "post_date": "2021-06-19T23:55:10.707000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I have read the discussions about all the possible data leakages, but that only got me to overfit 0.999, not 1.000. Could anyone share any insights? Just as a learning exercise, since we know the LB will be reset when the new data arrives.</p>",
  "messages": [
    {
      "id": 1357777,
      "postDate": "2021-06-20T00:08:44.500Z",
      "content": "<p>You may check the public kernel <a href=\"https://www.kaggle.com/darknesszx/leak-submission-lb-1-0/\" target=\"_blank\">here</a><br>\non how you can use the file's stats (last modification time), i.e. \"the leakage\" to classify with 100% accuracy </p>\n<p>check also  <a href=\"https://www.kaggle.com/c/seti-breakthrough-listen/discussion/247173\" target=\"_blank\">cpmp's comment</a> that explains very well why this happens - </p>",
      "rawMarkdown": "You may check the public kernel [here](https://www.kaggle.com/darknesszx/leak-submission-lb-1-0/)\non how you can use the file's stats (last modification time), i.e. \"the leakage\" to classify with 100% accuracy \n\ncheck also  [cpmp's comment](https://www.kaggle.com/c/seti-breakthrough-listen/discussion/247173) that explains very well why this happens - ",
      "votes": 3
    },
    {
      "id": 1357771,
      "postDate": "2021-06-19T23:55:10.707Z",
      "content": "<p>I have read the discussions about all the possible data leakages, but that only got me to overfit 0.999, not 1.000. Could anyone share any insights? Just as a learning exercise, since we know the LB will be reset when the new data arrives.</p>",
      "rawMarkdown": "I have read the discussions about all the possible data leakages, but that only got me to overfit 0.999, not 1.000. Could anyone share any insights? Just as a learning exercise, since we know the LB will be reset when the new data arrives."
    }
  ],
  "comments": [
    {
      "id": 1357777,
      "author_name": "Ioannis M",
      "author_url": "",
      "post_date": "2021-06-20T00:08:44.500000",
      "content": "<p>You may check the public kernel <a href=\"https://www.kaggle.com/darknesszx/leak-submission-lb-1-0/\" target=\"_blank\">here</a><br>\non how you can use the file's stats (last modification time), i.e. \"the leakage\" to classify with 100% accuracy </p>\n<p>check also  <a href=\"https://www.kaggle.com/c/seti-breakthrough-listen/discussion/247173\" target=\"_blank\">cpmp's comment</a> that explains very well why this happens - </p>",
      "votes": 3,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1357777": "You may check the public kernel [here](https://www.kaggle.com/darknesszx/leak-submission-lb-1-0/)\non how you can use the file's stats (last modification time), i.e. \"the leakage\" to classify with 100% accuracy \n\ncheck also  [cpmp's comment](https://www.kaggle.com/c/seti-breakthrough-listen/discussion/247173) that explains very well why this happens - ",
    "1357771": "I have read the discussions about all the possible data leakages, but that only got me to overfit 0.999, not 1.000. Could anyone share any insights? Just as a learning exercise, since we know the LB will be reset when the new data arrives."
  }
}