\n
![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F14573416%2Fa7fa973252d5eefcf5442abc147ded71%2FScreenshot%202024-03-18%20at%208.55.17PM.png?generation=1710791797556544&alt=media\")
Hi,
\nCurrent file on Kaggle:
\n
Local file (downloaded 21 of feb):
\n![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F14573416%2F1ab0adf488310265ee4a16ed75214644%2FScreenshot%202024-03-18%20at%208.58.47PM.png?generation=1710791959162240&alt=media\")
sha256sum:
\nKaggle file: 40db6c7c2514356224de81b6ffa89e8a58b94e84b479382880bc234a7e846c07
\nLocal file: 0501ce4d9eaec910cee3ea2e199feb79c5c99c19d90574d1c9754f151d271e78
Not clue what happened here, more values in general but also different ones for the same ids, 'contractssum_5085716L' isn't date relate so it shouldn't have changed.
", "messages": [ { "id": "2704608", "postDate": "03/18/2024 20:15:18", "content": "Hi,
\nCurrent file on Kaggle:
\n
Local file (downloaded 21 of feb):
\n
sha256sum:
\nKaggle file: 40db6c7c2514356224de81b6ffa89e8a58b94e84b479382880bc234a7e846c07
\nLocal file: 0501ce4d9eaec910cee3ea2e199feb79c5c99c19d90574d1c9754f151d271e78
Not clue what happened here, more values in general but also different ones for the same ids, 'contractssum_5085716L' isn't date relate so it shouldn't have changed.
", "rawMarkdown": "Hi,\n\nCurrent file on Kaggle:\n\n\nLocal file (downloaded 21 of feb):\n\n\nsha256sum:\nKaggle file: 40db6c7c2514356224de81b6ffa89e8a58b94e84b479382880bc234a7e846c07\nLocal file: 0501ce4d9eaec910cee3ea2e199feb79c5c99c19d90574d1c9754f151d271e78\n\nNot clue what happened here, more values in general but also different ones for the same ids, 'contractssum_5085716L' isn't date relate so it shouldn't have changed.", "votes": null }, { "id": "2705677", "postDate": "03/19/2024 13:13:22", "content": "When I asked a similar question I got that no transformation can be disclosed.
\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.
Yet, I got confirmation train is unchanged.
", "rawMarkdown": "When I asked a similar question I got that no transformation can be disclosed.\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.\n\nYet, I got confirmation train is unchanged.", "votes": null }, { "id": "2705907", "postDate": "03/19/2024 15:15:22", "content": "I believe you are referring to test data that are provided as test sample for testing. It is expected that there will be a difference as 1) there was a transformation of test set 2) case_ids are different.
", "rawMarkdown": "I believe you are referring to test data that are provided as test sample for testing. It is expected that there will be a difference as 1) there was a transformation of test set 2) case_ids are different.", "votes": null }, { "id": "2705917", "postDate": "03/19/2024 15:17:34", "content": "\n\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.
\n
We were considering this option, but 1) it doesn't represent the model running in production and as it is now it represents the reality more than the case when you would receive the date decision with case_id 2) it would disrupt even more the whole competition, kagglers already did a significant amount of work on that training data and we don't want to force everyone to start from scratch.
", "rawMarkdown": "> I don't get it, then just apply those \"unknown transformation\" to the Train set too.\n\nWe were considering this option, but 1) it doesn't represent the model running in production and as it is now it represents the reality more than the case when you would receive the date decision with case_id 2) it would disrupt even more the whole competition, kagglers already did a significant amount of work on that training data and we don't want to force everyone to start from scratch.", "votes": null }, { "id": "2705947", "postDate": "03/19/2024 15:32:12", "content": "The case_ids are the same, I know about the transformation, but columns that aren't relate to the dates also changed, more values plus different values, that's why I made the post, not clue if this was expected.
", "rawMarkdown": "The case_ids are the same, I know about the transformation, but columns that aren't relate to the dates also changed, more values plus different values, that's why I made the post, not clue if this was expected.", "votes": null }, { "id": "2705950", "postDate": "03/19/2024 15:33:45", "content": "The train dataset is equal, I tested it with polars.
\n![](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F14573416%2Ffb1dc492ae1e0b7f1fb2f2289ac06de2%2FScreenshot%202024-03-19%20at%204.28.03PM.png?generation=1710862376613293&alt=media\")
I think it is a very valid observation for 'contractssum_5085716L' e.g. case id 57631 the value changed. Case id 57630 the old value was null. These changes are not explained by dates transformation.
", "rawMarkdown": "I think it is a very valid observation for 'contractssum_5085716L' e.g. case id 57631 the value changed. Case id 57630 the old value was null. These changes are not explained by dates transformation.", "votes": null } ], "comments": [ { "id": 2705677, "author_name": "carloshuertas", "author_url": "", "post_date": "03/19/2024 13:13:22", "content": "When I asked a similar question I got that no transformation can be disclosed.
\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.
Yet, I got confirmation train is unchanged.
", "votes": null, "replies": [ { "id": 2705917, "author_name": "jetakow", "author_url": "", "post_date": "03/19/2024 15:17:34", "content": "\n\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.
\n
We were considering this option, but 1) it doesn't represent the model running in production and as it is now it represents the reality more than the case when you would receive the date decision with case_id 2) it would disrupt even more the whole competition, kagglers already did a significant amount of work on that training data and we don't want to force everyone to start from scratch.
", "votes": null, "replies": [] }, { "id": 2705950, "author_name": "enriquezaf", "author_url": "", "post_date": "03/19/2024 15:33:45", "content": "The train dataset is equal, I tested it with polars.
\n
I believe you are referring to test data that are provided as test sample for testing. It is expected that there will be a difference as 1) there was a transformation of test set 2) case_ids are different.
", "votes": null, "replies": [ { "id": 2705947, "author_name": "enriquezaf", "author_url": "", "post_date": "03/19/2024 15:32:12", "content": "The case_ids are the same, I know about the transformation, but columns that aren't relate to the dates also changed, more values plus different values, that's why I made the post, not clue if this was expected.
", "votes": null, "replies": [] }, { "id": 2706528, "author_name": "fadylabib", "author_url": "", "post_date": "03/19/2024 23:22:42", "content": "I think it is a very valid observation for 'contractssum_5085716L' e.g. case id 57631 the value changed. Case id 57630 the old value was null. These changes are not explained by dates transformation.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "2704608": "Hi,\n\nCurrent file on Kaggle:\n\n\nLocal file (downloaded 21 of feb):\n\n\nsha256sum:\nKaggle file: 40db6c7c2514356224de81b6ffa89e8a58b94e84b479382880bc234a7e846c07\nLocal file: 0501ce4d9eaec910cee3ea2e199feb79c5c99c19d90574d1c9754f151d271e78\n\nNot clue what happened here, more values in general but also different ones for the same ids, 'contractssum_5085716L' isn't date relate so it shouldn't have changed.", "2705677": "When I asked a similar question I got that no transformation can be disclosed.\nI don't get it, then just apply those \"unknown transformation\" to the Train set too.\n\nYet, I got confirmation train is unchanged.", "2705907": "I believe you are referring to test data that are provided as test sample for testing. It is expected that there will be a difference as 1) there was a transformation of test set 2) case_ids are different.", "2705917": "> I don't get it, then just apply those \"unknown transformation\" to the Train set too.\n\nWe were considering this option, but 1) it doesn't represent the model running in production and as it is now it represents the reality more than the case when you would receive the date decision with case_id 2) it would disrupt even more the whole competition, kagglers already did a significant amount of work on that training data and we don't want to force everyone to start from scratch.", "2705947": "The case_ids are the same, I know about the transformation, but columns that aren't relate to the dates also changed, more values plus different values, that's why I made the post, not clue if this was expected.", "2705950": "The train dataset is equal, I tested it with polars.\n\n", "2706528": "I think it is a very valid observation for 'contractssum_5085716L' e.g. case id 57631 the value changed. Case id 57630 the old value was null. These changes are not explained by dates transformation." }, "source": "meta" }