This competition is unique in that it uses the Competition API to create a real-world situation.
\nThe API does not allow the participant to know the test data when training the model. Compared to usual Kaggle competition, it's much closer to real-world problems.
\nAt first, I had some trouble understanding the API at first, but host's notebooks are really helpful.
\nhttps://www.kaggle.com/sohier/quick-sample-submission
\nhttps://www.kaggle.com/sohier/competition-api-detailed-introduction
And I published the benchmark notebook which includes the process of LGBM with 5 cross-validation. I hope this help participants understand the task.
\nhttps://www.kaggle.com/sishihara/riiid-lgbm-5cv-benchmark
This competition is unique in that it uses the Competition API to create a real-world situation.
\nThe API does not allow the participant to know the test data when training the model. Compared to usual Kaggle competition, it's much closer to real-world problems.
\nAt first, I had some trouble understanding the API at first, but host's notebooks are really helpful.
\nhttps://www.kaggle.com/sohier/quick-sample-submission
\nhttps://www.kaggle.com/sohier/competition-api-detailed-introduction
And I published the benchmark notebook which includes the process of LGBM with 5 cross-validation. I hope this help participants understand the task.
\nhttps://www.kaggle.com/sishihara/riiid-lgbm-5cv-benchmark
HEY,i can understand that dude but some way or another it is one of the best Program with best deliverance
", "rawMarkdown": "HEY,i can understand that dude but some way or another it is one of the best Program with best deliverance", "votes": null }, { "id": "1039057", "postDate": "10/06/2020 09:38:37", "content": "this i s one of the best thing i come across to play with
", "rawMarkdown": "this i s one of the best thing i come across to play with", "votes": null }, { "id": "1039375", "postDate": "10/06/2020 14:34:47", "content": "Thanks.
", "rawMarkdown": "Thanks.", "votes": null }, { "id": "1044334", "postDate": "10/09/2020 17:13:01", "content": "Isn't it possible that people can collect the test data in a global list outside and see the sample at-least that's available? Apologies, new to this comp.
\nStreaming data this way is also an issue as we cannot control the batch_size nor can we calculate aggs easily as one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch..
", "rawMarkdown": "Isn't it possible that people can collect the test data in a global list outside and see the sample at-least that's available? Apologies, new to this comp.\n\nStreaming data this way is also an issue as we cannot control the batch_size nor can we calculate aggs easily as one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch..", "votes": null }, { "id": "1044374", "postDate": "10/09/2020 17:48:49", "content": "\n\none user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch.
\n
Which is precisely how it would occur in real world and hence the competition is setup this way to replicate it as closely as possible.
", "rawMarkdown": "> one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch.\n\nWhich is precisely how it would occur in real world and hence the competition is setup this way to replicate it as closely as possible.", "votes": null }, { "id": "1044389", "postDate": "10/09/2020 18:03:52", "content": "Yep I agree to that Rohan! The comp's gonna be fun as now there's a good % in the comp that demands good SDE skills as well :)
", "rawMarkdown": "Yep I agree to that Rohan! The comp's gonna be fun as now there's a good % in the comp that demands good SDE skills as well :)", "votes": null }, { "id": "1044734", "postDate": "10/10/2020 04:43:23", "content": "@sohier @jineonjinbaek
\n1) Is the public test data only have that 104 rows or is it just an example data?
\n2) why only some rows in \"prior_group_response\" have values in the form of list and other rows have null values?
\n![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F3307250%2F40b0ffa9ecf5d3bf9f2067307a3206d9%2Fq.png?generation=1602305818072160&alt=media\")
\n3) If you are using Python, you will likely want to call \"eval\" on the non-null rows. Some rows may be null, or empty lists. What did the competition host meant by this?
Could anyone shed some light on these question. It would be really helpful for me to get started in this competition.
", "rawMarkdown": "sohier @jineonjinbaek \n1) Is the public test data only have that 104 rows or is it just an example data?\n2) why only some rows in \"prior_group_response\" have values in the form of list and other rows have null values?\n\n3) `If you are using Python, you will likely want to call \"eval\" on the non-null rows. Some rows may be null, or empty lists.` What did the competition host meant by this?\n\nCould anyone shed some light on these question. It would be really helpful for me to get started in this competition.", "votes": null }, { "id": "1044750", "postDate": "10/10/2020 04:57:41", "content": "@adityaecdrid thank you for your replies
\nBut you answered for the 1st and 3rd question.
\nDo you know answer of my 2nd question??
in description
\n\n", "rawMarkdown": "in [description](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n> prior_group_responses (string) provides all of the user_answer entries for previous group in a string representation of a list **in the first row of the group**. All other rows in each group are null.", "votes": null }, { "id": "1101586", "postDate": "12/04/2020 03:39:40", "content": "prior_group_responses (string) provides all of the user_answer entries for previous group in a string representation of a list in the first row of the group. All other rows in each group are null.
\n
So I still do not know why I cannot collect it in a global list. Do you know it now?
", "rawMarkdown": "So I still do not know why I cannot collect it in a global list. Do you know it now?", "votes": null } ], "comments": [ { "id": 1039057, "author_name": "ayusharma18", "author_url": "", "post_date": "10/06/2020 09:38:37", "content": "this i s one of the best thing i come across to play with
", "votes": null, "replies": [] }, { "id": 1039375, "author_name": "cdeotte", "author_url": "", "post_date": "10/06/2020 14:34:47", "content": "Thanks.
", "votes": null, "replies": [] }, { "id": 1044334, "author_name": "adityaecdrid", "author_url": "", "post_date": "10/09/2020 17:13:01", "content": "Isn't it possible that people can collect the test data in a global list outside and see the sample at-least that's available? Apologies, new to this comp.
\nStreaming data this way is also an issue as we cannot control the batch_size nor can we calculate aggs easily as one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch..
", "votes": null, "replies": [ { "id": 1044374, "author_name": "rohanrao", "author_url": "", "post_date": "10/09/2020 17:48:49", "content": "\n\none user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch.
\n
Which is precisely how it would occur in real world and hence the competition is setup this way to replicate it as closely as possible.
", "votes": null, "replies": [] }, { "id": 1044389, "author_name": "adityaecdrid", "author_url": "", "post_date": "10/09/2020 18:03:52", "content": "Yep I agree to that Rohan! The comp's gonna be fun as now there's a good % in the comp that demands good SDE skills as well :)
", "votes": null, "replies": [] }, { "id": 1101586, "author_name": "weikangqiu", "author_url": "", "post_date": "12/04/2020 03:39:40", "content": "So I still do not know why I cannot collect it in a global list. Do you know it now?
", "votes": null, "replies": [] } ] }, { "id": 1044734, "author_name": "msharuk589", "author_url": "", "post_date": "10/10/2020 04:43:23", "content": "@sohier @jineonjinbaek
\n1) Is the public test data only have that 104 rows or is it just an example data?
\n2) why only some rows in \"prior_group_response\" have values in the form of list and other rows have null values?
\n
\n3) If you are using Python, you will likely want to call \"eval\" on the non-null rows. Some rows may be null, or empty lists. What did the competition host meant by this?
Could anyone shed some light on these question. It would be really helpful for me to get started in this competition.
", "votes": null, "replies": [ { "id": 1044750, "author_name": "adityaecdrid", "author_url": "", "post_date": "10/10/2020 04:57:41", "content": "@adityaecdrid thank you for your replies
\nBut you answered for the 1st and 3rd question.
\nDo you know answer of my 2nd question??
in description
\n\n", "votes": null, "replies": [] } ] }, { "id": 1039045, "author_name": "ayusharma18", "author_url": "", "post_date": "10/06/2020 09:30:21", "content": "prior_group_responses (string) provides all of the user_answer entries for previous group in a string representation of a list in the first row of the group. All other rows in each group are null.
\n
HEY,i can understand that dude but some way or another it is one of the best Program with best deliverance
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1038435": "This competition is unique in that it uses the Competition API to create a real-world situation. \n\nThe API does not allow the participant to know the test data when training the model. Compared to usual Kaggle competition, it's much closer to real-world problems.\n\nAt first, I had some trouble understanding the API at first, but host's notebooks are really helpful.\nhttps://www.kaggle.com/sohier/quick-sample-submission\nhttps://www.kaggle.com/sohier/competition-api-detailed-introduction\n\nAnd I published the benchmark notebook which includes the process of LGBM with 5 cross-validation. I hope this help participants understand the task.\nhttps://www.kaggle.com/sishihara/riiid-lgbm-5cv-benchmark", "1039045": "HEY,i can understand that dude but some way or another it is one of the best Program with best deliverance", "1039057": "this i s one of the best thing i come across to play with", "1039375": "Thanks.", "1044334": "Isn't it possible that people can collect the test data in a global list outside and see the sample at-least that's available? Apologies, new to this comp.\n\nStreaming data this way is also an issue as we cannot control the batch_size nor can we calculate aggs easily as one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch..", "1044374": "> one user can appear in kth batch and the same user will again appear in k+10th batch etc? We don't have a guarantee that all user have to appear in a single batch.\n\nWhich is precisely how it would occur in real world and hence the competition is setup this way to replicate it as closely as possible.", "1044389": "Yep I agree to that Rohan! The comp's gonna be fun as now there's a good % in the comp that demands good SDE skills as well :)", "1044734": "sohier @jineonjinbaek \n1) Is the public test data only have that 104 rows or is it just an example data?\n2) why only some rows in \"prior_group_response\" have values in the form of list and other rows have null values?\n\n3) `If you are using Python, you will likely want to call \"eval\" on the non-null rows. Some rows may be null, or empty lists.` What did the competition host meant by this?\n\nCould anyone shed some light on these question. It would be really helpful for me to get started in this competition.", "1044750": "1. NO obviously. (you have ~ 2.5 M rows in test set) Consider reading the data description over and over unless you got all the points implanted in your memory!\n3. You have to convert that string which is actually a list, that's why you use eval (google up ast's literal eval in python)", "1044841": "adityaecdrid thank you for your replies\nBut you answered for the 1st and 3rd question.\nDo you know answer of my 2nd question??", "1044937": "in [description](https://www.kaggle.com/c/riiid-test-answer-prediction/data)\n> prior_group_responses (string) provides all of the user_answer entries for previous group in a string representation of a list **in the first row of the group**. All other rows in each group are null.", "1101586": "So I still do not know why I cannot collect it in a global list. Do you know it now?" }, "source": "meta" }