{ "id": 501853, "title": "Is the WEEK_NUM in sample_submission.csv fake?", "url": "/competitions/home-credit-credit-risk-model-stability/discussion/501853", "author_name": "Bruce", "post_date": "2024-05-11T01:32:27.676000", "votes": 6, "comment_count": 4, "views": 0, "content": "

Has the official made any adjustments to the WEEK_NUM in sample_submission.csv?
\nIn that case, using the previous Magic trick on prediction results won't improve anymore?
\nTherefore, if one wants to hack the metric, should they obtain the correct WEEK_NUM first?

", "messages": [ { "id": 2806251, "postDate": "2024-05-11T01:32:27.677Z", "content": "

Has the official made any adjustments to the WEEK_NUM in sample_submission.csv?
\nIn that case, using the previous Magic trick on prediction results won't improve anymore?
\nTherefore, if one wants to hack the metric, should they obtain the correct WEEK_NUM first?

", "rawMarkdown": "Has the official made any adjustments to the WEEK_NUM in sample_submission.csv?\nIn that case, using the previous Magic trick on prediction results won't improve anymore? \nTherefore, if one wants to hack the metric, should they obtain the correct WEEK_NUM first?", "votes": 6 }, { "id": 2807101, "postDate": "2024-05-11T13:52:31.663Z", "content": "

Want to confirm, so in the test_base.csv, the only real data are case_id and date_decision, and the other ones (WEEK_NUM and MONTH) are masked? Thanks!

", "rawMarkdown": "Want to confirm, so in the test_base.csv, the only real data are **case_id** and **date_decision**, and the other ones (**WEEK_NUM** and **MONTH**) are masked? Thanks!", "votes": 1, "replies": [ { "id": 2810448, "postDate": "2024-05-13T09:19:02.237Z", "content": "

Hi @faithk7u ,

\n

here is our statement when submissions have been re-opened (https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474):

\n
\n

As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value. We decided to keep them to minimize the amount of time you'll need to spend altering your scripts to get the submissions working again. Column date_decision is no longer the same. The remaining data were also transformed.

\n
", "rawMarkdown": "Hi @faithk7u ,\n\nhere is our statement when submissions have been re-opened (https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474):\n\n>As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value. We decided to keep them to minimize the amount of time you'll need to spend altering your scripts to get the submissions working again. Column date_decision is no longer the same. The remaining data were also transformed.", "votes": 1, "replies": [ { "id": 2810990, "postDate": "2024-05-13T14:54:05.157Z", "rawMarkdown": "", "isDeleted": true } ] } ] }, { "id": 2806424, "postDate": "2024-05-11T04:58:58.343Z", "content": "

The sample submission is just 10 rows of data to test your code. The week num column is fake here. But from other posts, we are clear that one could reasonably estimate this for the test (actual) data @bruceqdu

", "rawMarkdown": "The sample submission is just 10 rows of data to test your code. The week num column is fake here. But from other posts, we are clear that one could reasonably estimate this for the test (actual) data @bruceqdu " } ], "comments": [ { "id": 2807101, "author_name": "faithk7u", "author_url": "", "post_date": "2024-05-11T13:52:31.663000", "content": "

Want to confirm, so in the test_base.csv, the only real data are case_id and date_decision, and the other ones (WEEK_NUM and MONTH) are masked? Thanks!

", "votes": 1, "replies": [ { "id": 2810448, "author_name": "Tomas Jelinek", "author_url": "", "post_date": "2024-05-13T09:19:02.237000", "content": "

Hi @faithk7u ,

\n

here is our statement when submissions have been re-opened (https://www.kaggle.com/competitions/home-credit-credit-risk-model-stability/discussion/482474):

\n
\n

As we indicated previously, we will be changing date columns in test_base table, namely date_decision, MONTH and WEEK_NUM. The MONTH and WEEK_NUM columns will now only have one constant value. We decided to keep them to minimize the amount of time you'll need to spend altering your scripts to get the submissions working again. Column date_decision is no longer the same. The remaining data were also transformed.

\n
", "votes": 1, "replies": [ { "id": 2810990, "author_name": "", "author_url": "", "post_date": "2024-05-13T14:54:05.157000", "content": "", "votes": 0, "replies": [] } ] } ] }, { "id": 2806424, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2024-05-11T04:58:58.343000", "content": "

The sample submission is just 10 rows of data to test your code. The week num column is fake here. But from other posts, we are clear that one could reasonably estimate this for the test (actual) data @bruceqdu

", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2806251": "Has the official made any adjustments to the WEEK_NUM in sample_submission.csv?\nIn that case, using the previous Magic trick on prediction results won't improve anymore? \nTherefore, if one wants to hack the metric, should they obtain the correct WEEK_NUM first?", "2807101": "Want to confirm, so in the test_base.csv, the only real data are **case_id** and **date_decision**, and the other ones (**WEEK_NUM** and **MONTH**) are masked? Thanks!", "2806424": "The sample submission is just 10 rows of data to test your code. The week num column is fake here. But from other posts, we are clear that one could reasonably estimate this for the test (actual) data @bruceqdu " } }