I found some logical bugs in my code.
\nIt comes from
\ndata.groupby(['user_id', 'task_container_id'])
\ndata.groupby(['user_id', 'timestamp '])
I think the reason is
\nfor example
\n| timestamp | \nuser_id | \ntask_container_id | \nanswered_correctly | \n
|---|---|---|---|
| 166938177 | \n712889458 | \n57 | \n1 | \n
| 167006365 | \n712889458 | \n59 | \n1 | \n
| 167006365 | \n712889458 | \n58 | \n0 | \n
| 167006365 | \n712889458 | \n60 | \n1 | \n
It puzzled me until I find this topic
\nIf it's new to you, hope you get some tips from here
I found some logical bugs in my code.
\nIt comes from
\ndata.groupby(['user_id', 'task_container_id'])
\ndata.groupby(['user_id', 'timestamp '])
I think the reason is
\nfor example
\n| timestamp | \nuser_id | \ntask_container_id | \nanswered_correctly | \n
|---|---|---|---|
| 166938177 | \n712889458 | \n57 | \n1 | \n
| 167006365 | \n712889458 | \n59 | \n1 | \n
| 167006365 | \n712889458 | \n58 | \n0 | \n
| 167006365 | \n712889458 | \n60 | \n1 | \n
It puzzled me until I find this topic
\nIf it's new to you, hope you get some tips from here
Why is the timestamp the same but the task_container_id is different?
\nIn my opinion, if the timestamp is the same, the questions are served together and task_container_id will be the same.
\n\n", "rawMarkdown": "> The timestamp column shows when an activity is finished, not when it started. A user could have spent an arbitrary amount of time working on the first problem before the first timestamp value would get logged.\n\n[more details](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465)", "votes": null }, { "id": "1099138", "postDate": "12/02/2020 05:38:08", "content": "The timestamp column shows when an activity is finished, not when it started. A user could have spent an arbitrary amount of time working on the first problem before the first timestamp value would get logged.
\n
Simply add sort=False in groupby function and it should work fine.
You're right, dude
\nBut I did not know these timestap and task_container_id information before
Hi Solomon, Sorry for the n00b question.
\nHow is this a logical bug? It looks like consistently use task_container_id or timestamp should be fine.
\nAlso, based on my understanding, add sort=False in groupby function will just make the result unsorted, what's the special need here?
\nAppreciate any educating here
Because I did not know
\nI created features base on them which I thought were in order,
\nand of cause I did not recive error or warning.
\nIt means I need to find these bugs by myself. (Even more, I need to realise it at first)
\nYou may call it \"careless bug\"😃
Thank you for the quick reply. I see, so your feature rely on the order and it's more like a logic bug?
\nAlso, as I learned from description of API,
\n\nEach group will contain interactions from many different users, but no more than one task_container_id of questions from any single user.
\n
which means during testing phase in each group, each user will only have 1 single task_container_id. Do I understand correctly?
", "rawMarkdown": "Thank you for the quick reply. I see, so your feature rely on the order and it's more like a logic bug?\nAlso, as I learned from description of API, \n> Each group will contain interactions from many different users, but no more than one task_container_id of questions from any single user. \n\nwhich means during testing phase in each group, each user will only have 1 single task_container_id. Do I understand correctly?", "votes": null }, { "id": "1100458", "postDate": "12/03/2020 05:13:13", "content": "You're right
\nJust remember timestamp is the finish time, and task id is not totally in order
This are the only tuples that have repeated task_id I do not think it is much of a problem
\nuser_id timestamp
\n7171715 6828386 2
\n37317993 1729136465 2
\n169835132 16435464437 2
\n575358140 2136919285 2
\n712889458 167006365 2
\n743195685 2014787948 2
\n828365782 6067874053 2
\n857524087 11647087 2
\n917426026 170401915 2
\n993524405 1351161810 2
But part 1 is surely strange.
", "rawMarkdown": "This are the only tuples that have repeated task_id I do not think it is much of a problem\nuser_id timestamp \n7171715 6828386 2\n37317993 1729136465 2\n169835132 16435464437 2\n575358140 2136919285 2\n712889458 167006365 2\n743195685 2014787948 2\n828365782 6067874053 2\n857524087 11647087 2\n917426026 170401915 2\n993524405 1351161810 2\n\nBut part 1 is surely strange.", "votes": null }, { "id": "1101730", "postDate": "12/04/2020 07:20:16", "content": "Can a user run two or more sessions simultaneously?
", "rawMarkdown": "Can a user run two or more sessions simultaneously?", "votes": null }, { "id": "1101765", "postDate": "12/04/2020 08:14:37", "content": "It seems Yes.
\nIn my opinion, task id shows the start order, and timestamp shows the finish order
Why is the timestamp the same but the task_container_id is different?
\nIn my opinion, if the timestamp is the same, the questions are served together and task_container_id will be the same.
\n\n", "votes": null, "replies": [] }, { "id": 1101730, "author_name": "neustroev", "author_url": "", "post_date": "12/04/2020 07:20:16", "content": "The timestamp column shows when an activity is finished, not when it started. A user could have spent an arbitrary amount of time working on the first problem before the first timestamp value would get logged.
\n
Can a user run two or more sessions simultaneously?
", "votes": null, "replies": [] }, { "id": 1101765, "author_name": "solomonxian", "author_url": "", "post_date": "12/04/2020 08:14:37", "content": "It seems Yes.
\nIn my opinion, task id shows the start order, and timestamp shows the finish order
Simply add sort=False in groupby function and it should work fine.
You're right, dude
\nBut I did not know these timestap and task_container_id information before
Hi Solomon, Sorry for the n00b question.
\nHow is this a logical bug? It looks like consistently use task_container_id or timestamp should be fine.
\nAlso, based on my understanding, add sort=False in groupby function will just make the result unsorted, what's the special need here?
\nAppreciate any educating here
Because I did not know
\nI created features base on them which I thought were in order,
\nand of cause I did not recive error or warning.
\nIt means I need to find these bugs by myself. (Even more, I need to realise it at first)
\nYou may call it \"careless bug\"😃
Thank you for the quick reply. I see, so your feature rely on the order and it's more like a logic bug?
\nAlso, as I learned from description of API,
\n\nEach group will contain interactions from many different users, but no more than one task_container_id of questions from any single user.
\n
which means during testing phase in each group, each user will only have 1 single task_container_id. Do I understand correctly?
", "votes": null, "replies": [] }, { "id": 1100458, "author_name": "solomonxian", "author_url": "", "post_date": "12/03/2020 05:13:13", "content": "You're right
\nJust remember timestamp is the finish time, and task id is not totally in order
This are the only tuples that have repeated task_id I do not think it is much of a problem
\nuser_id timestamp
\n7171715 6828386 2
\n37317993 1729136465 2
\n169835132 16435464437 2
\n575358140 2136919285 2
\n712889458 167006365 2
\n743195685 2014787948 2
\n828365782 6067874053 2
\n857524087 11647087 2
\n917426026 170401915 2
\n993524405 1351161810 2
But part 1 is surely strange.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1097788": "I found some logical bugs in my code.\nIt comes from \n`data.groupby(['user_id', 'task_container_id'])` \n`data.groupby(['user_id', 'timestamp '])`\n\nI think the reason is \n1. task_container_id is not monotonically increasing for each user as [this topic describes](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189513)\n2. some tasks share the same timestamp \n\nfor example\n| timestamp | user_id | task_container_id | answered_correctly\n| --- | --- |\n| 166938177| 712889458|57 | 1\n| 167006365| 712889458|59| 1\n| 167006365| 712889458|58| 0\n| 167006365| 712889458|60| 1\n\nIt puzzled me until I find [this topic](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465)\nIf it's new to you, hope you get some tips from here", "1098979": "Why is the timestamp the same but the task_container_id is different?\nIn my opinion, if the timestamp is the same, the questions are served together and task_container_id will be the same.", "1099087": "> The timestamp column shows when an activity is finished, not when it started. A user could have spent an arbitrary amount of time working on the first problem before the first timestamp value would get logged.\n\n[more details](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465)", "1099138": "Simply add `sort=False` in groupby function and it should work fine.", "1099226": "You're right, dude\nBut I did not know these timestap and task_container_id information before", "1100373": "Hi Solomon, Sorry for the n00b question. \nHow is this a logical bug? It looks like consistently use task_container_id or timestamp should be fine.\nAlso, based on my understanding, add `sort=False` in groupby function will just make the result unsorted, what's the special need here?\nAppreciate any educating here", "1100419": "Because I did not know \n1. task_container_id is not monotonically increasing for each user\n2. timestamp is not monotonically increasing for each user\n\nI created features base on them which I thought were in order, \nand of cause I did not recive error or warning.\nIt means I need to find these bugs by myself. (Even more, I need to realise it at first)\nYou may call it \"careless bug\"😃", "1100433": "Thank you for the quick reply. I see, so your feature rely on the order and it's more like a logic bug?\nAlso, as I learned from description of API, \n> Each group will contain interactions from many different users, but no more than one task_container_id of questions from any single user. \n\nwhich means during testing phase in each group, each user will only have 1 single task_container_id. Do I understand correctly?", "1100458": "You're right \nJust remember timestamp is the finish time, and task id is not totally in order", "1100476": "This are the only tuples that have repeated task_id I do not think it is much of a problem\nuser_id timestamp \n7171715 6828386 2\n37317993 1729136465 2\n169835132 16435464437 2\n575358140 2136919285 2\n712889458 167006365 2\n743195685 2014787948 2\n828365782 6067874053 2\n857524087 11647087 2\n917426026 170401915 2\n993524405 1351161810 2\n\nBut part 1 is surely strange.", "1101730": "Can a user run two or more sessions simultaneously?", "1101765": "It seems Yes.\nIn my opinion, task id shows the start order, and timestamp shows the finish order" }, "source": "meta" }