{
  "id": 507963,
  "title": "private score 0.605 ran away",
  "url": "/competitions/home-credit-credit-risk-model-stability/discussion/507963",
  "author_name": "",
  "post_date": "2024-05-28T02:02:09.922809Z",
  "votes": 5,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I succeeded in pseudo-recovering the month of the test data, but I couldn't get a good score because I set the period during which the score was negative to one quarter of the max_day minus min_day.<br>\nHowever, when I looked at my other submissions, I found that I had obtained a score of 0.605 by setting the period during which the score was to be reduced to 1/2 of the part obtained by subtracting min_day from max_day.<br>\nThis is probably due to the private test data chronologically following the end of the public test data.</p>\n<p>This result will be the subject of my funny story (lol)</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20605011%2F877f37611682362ef8437ebb8ec2b0d4%2Fprivate_score.png?generation=1716861459650679&amp;alt=media\"></p>\n<pre><code>\nmax_day = ([i  i  df_subm[]  i   ])\nmin_day = ([i  i  df_subm[]  i   ])\nmin_day = ([min_day,pd.Timestamp(date(,,))])\nfirst_half_period = (max_day - min_day)//\n\ncondition = df_subm[]     df_subm[] &lt; first_half_period + min_day\ndf_subm.loc[condition, ] = (df_subm.loc[condition, ] - ).clip()\n</code></pre>\n<p>Metrics used to restore the month<br>\npmts_dpd_1073P,pmts_month_158T,pmts_year_1139T</p>",
  "messages": [
    {
      "id": "2840160",
      "postDate": "05/28/2024 02:02:09",
      "content": "<p>I succeeded in pseudo-recovering the month of the test data, but I couldn't get a good score because I set the period during which the score was negative to one quarter of the max_day minus min_day.<br>\nHowever, when I looked at my other submissions, I found that I had obtained a score of 0.605 by setting the period during which the score was to be reduced to 1/2 of the part obtained by subtracting min_day from max_day.<br>\nThis is probably due to the private test data chronologically following the end of the public test data.</p>\n<p>This result will be the subject of my funny story (lol)</p>\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20605011%2F877f37611682362ef8437ebb8ec2b0d4%2Fprivate_score.png?generation=1716861459650679&amp;alt=media\"></p>\n<pre><code>\nmax_day = ([i  i  df_subm[]  i   ])\nmin_day = ([i  i  df_subm[]  i   ])\nmin_day = ([min_day,pd.Timestamp(date(,,))])\nfirst_half_period = (max_day - min_day)//\n\ncondition = df_subm[]     df_subm[] &lt; first_half_period + min_day\ndf_subm.loc[condition, ] = (df_subm.loc[condition, ] - ).clip()\n</code></pre>\n<p>Metrics used to restore the month<br>\npmts_dpd_1073P,pmts_month_158T,pmts_year_1139T</p>",
      "rawMarkdown": "I succeeded in pseudo-recovering the month of the test data, but I couldn't get a good score because I set the period during which the score was negative to one quarter of the max_day minus min_day.\nHowever, when I looked at my other submissions, I found that I had obtained a score of 0.605 by setting the period during which the score was to be reduced to 1/2 of the part obtained by subtracting min_day from max_day.\nThis is probably due to the private test data chronologically following the end of the public test data.\n\nThis result will be the subject of my funny story (lol)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20605011%2F877f37611682362ef8437ebb8ec2b0d4%2Fprivate_score.png?generation=1716861459650679&alt=media)\n\n```python\n#max_payment_day is reconstruction of the month\nmax_day = max([i for i in df_subm[\"max_payment_day\"] if i is not None])\nmin_day = min([i for i in df_subm[\"max_payment_day\"] if i is not None])\nmin_day = max([min_day,pd.Timestamp(date(2020,10,6))])\nfirst_half_period = (max_day - min_day)//2\n\ncondition = df_subm[\"max_payment_day\"] is not None and df_subm[\"max_payment_day\"] < first_half_period + min_day\ndf_subm.loc[condition, 'score'] = (df_subm.loc[condition, 'score'] - 0.045).clip(0)\n```\nMetrics used to restore the month\npmts_dpd_1073P,pmts_month_158T,pmts_year_1139T",
      "votes": null
    },
    {
      "id": "2840724",
      "postDate": "05/28/2024 08:29:46",
      "content": "<p>Life is a joke lol</p>",
      "rawMarkdown": "Life is a joke lol",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2840724,
      "author_name": "edenwild",
      "author_url": "",
      "post_date": "05/28/2024 08:29:46",
      "content": "<p>Life is a joke lol</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2840160": "I succeeded in pseudo-recovering the month of the test data, but I couldn't get a good score because I set the period during which the score was negative to one quarter of the max_day minus min_day.\nHowever, when I looked at my other submissions, I found that I had obtained a score of 0.605 by setting the period during which the score was to be reduced to 1/2 of the part obtained by subtracting min_day from max_day.\nThis is probably due to the private test data chronologically following the end of the public test data.\n\nThis result will be the subject of my funny story (lol)\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F20605011%2F877f37611682362ef8437ebb8ec2b0d4%2Fprivate_score.png?generation=1716861459650679&alt=media)\n\n```python\n#max_payment_day is reconstruction of the month\nmax_day = max([i for i in df_subm[\"max_payment_day\"] if i is not None])\nmin_day = min([i for i in df_subm[\"max_payment_day\"] if i is not None])\nmin_day = max([min_day,pd.Timestamp(date(2020,10,6))])\nfirst_half_period = (max_day - min_day)//2\n\ncondition = df_subm[\"max_payment_day\"] is not None and df_subm[\"max_payment_day\"] < first_half_period + min_day\ndf_subm.loc[condition, 'score'] = (df_subm.loc[condition, 'score'] - 0.045).clip(0)\n```\nMetrics used to restore the month\npmts_dpd_1073P,pmts_month_158T,pmts_year_1139T",
    "2840724": "Life is a joke lol"
  },
  "source": "meta"
}