{ "id": 193951, "title": "Multi tasking users? ", "url": "/competitions/riiid-test-answer-prediction/discussion/193951", "author_name": "", "post_date": "2020-10-29T20:23:57.778207300Z", "votes": 1, "comment_count": 2, "views": 0, "content": "

While exploring the train.csv data, I came across this finding. For 1M rows, the key points I noticed:

\n\n
    \n
  1. They have neighboring content IDs, e.g. 6787, 6788,6786,6785.
    2. \n
  2. They have exactly the same prior_question_elapsed_time and task_container_id.
    3. \n
\n

Can anyone give any explanation for this occurrence?
\nPS: I have explored these findings for timestamp==0 only. There might be further occurrences, but I couldn't find any explanation for this. Is it that the user is solving those questions or started solving the questions at once? Or the question bundle? But they have different yet neighboring content_id. Hence, do content_id preserves some sequential information?
\nHere is my notebook: https://www.kaggle.com/mrutyunjaybiswal/riiid-aied-challenge-2020-data-exploration-and-fe

", "messages": [ { "id": "1064189", "postDate": "10/29/2020 20:23:57", "content": "

While exploring the train.csv data, I came across this finding. For 1M rows, the key points I noticed:

\n\n
    \n
  1. They have neighboring content IDs, e.g. 6787, 6788,6786,6785.
    2. \n
  2. They have exactly the same prior_question_elapsed_time and task_container_id.
    3. \n
\n

Can anyone give any explanation for this occurrence?
\nPS: I have explored these findings for timestamp==0 only. There might be further occurrences, but I couldn't find any explanation for this. Is it that the user is solving those questions or started solving the questions at once? Or the question bundle? But they have different yet neighboring content_id. Hence, do content_id preserves some sequential information?
\nHere is my notebook: https://www.kaggle.com/mrutyunjaybiswal/riiid-aied-challenge-2020-data-exploration-and-fe

", "rawMarkdown": "While exploring the `train.csv` data, I came across this finding. For 1M rows, the key points I noticed:\n\n- There are 8 such users with multiple interactions at `timestamp == 0`.\n- The common thing among these users are:\n1. They have neighboring content IDs, e.g. 6787, 6788,6786,6785.\n2. They have exactly the same `prior_question_elapsed_time` and `task_container_id`.\n\nCan anyone give any explanation for this occurrence? \nPS: I have explored these findings for `timestamp==0` only. There might be further occurrences, but I couldn't find any explanation for this. Is it that the user is solving those questions or started solving the questions at once? Or the question bundle? But they have different yet neighboring `content_id`. Hence, do `content_id` preserves some sequential information?\nHere is my notebook: https://www.kaggle.com/mrutyunjaybiswal/riiid-aied-challenge-2020-data-exploration-and-fe", "votes": null }, { "id": "1064196", "postDate": "10/29/2020 20:34:24", "content": "

This is expected behavior, it just represents a bundle of questions. The timestamp indicates when a single task is completed relative to the first task completed, and a single task_container_id for a bundle shares its timestamp across all the individual questions that are part of that bundle. Prior_question_elapsed_time is also shared across the entire bundle. So in summary, these question bundles are totally normal and occur often in the data, you're just looking at the subset of them that occurred at the beginning of user history (which seems fairly uncommon, given your findings).

", "rawMarkdown": "This is expected behavior, it just represents a bundle of questions. The `timestamp` indicates when a single task is completed relative to the first task completed, and a single `task_container_id` for a bundle shares its timestamp across all the individual questions that are part of that bundle. `Prior_question_elapsed_time` is also shared across the entire bundle. So in summary, these question bundles are totally normal and occur often in the data, you're just looking at the subset of them that occurred at the beginning of user history (which seems fairly uncommon, given your findings).", "votes": null }, { "id": "1064304", "postDate": "10/30/2020 00:44:03", "content": "

Since these are all timestamp =0 interactions, prior_question_elapsed_time must be NA/0 for all. Multiple interactions at timestamp 0 imply that this user was served a bundle of questions on his first session.

", "rawMarkdown": "Since these are all timestamp =0 interactions, prior_question_elapsed_time must be NA/0 for all. Multiple interactions at timestamp 0 imply that this user was served a bundle of questions on his first session.", "votes": null } ], "comments": [ { "id": 1064196, "author_name": "aquatic", "author_url": "", "post_date": "10/29/2020 20:34:24", "content": "

This is expected behavior, it just represents a bundle of questions. The timestamp indicates when a single task is completed relative to the first task completed, and a single task_container_id for a bundle shares its timestamp across all the individual questions that are part of that bundle. Prior_question_elapsed_time is also shared across the entire bundle. So in summary, these question bundles are totally normal and occur often in the data, you're just looking at the subset of them that occurred at the beginning of user history (which seems fairly uncommon, given your findings).

", "votes": null, "replies": [] }, { "id": 1064304, "author_name": "abhimanyud", "author_url": "", "post_date": "10/30/2020 00:44:03", "content": "

Since these are all timestamp =0 interactions, prior_question_elapsed_time must be NA/0 for all. Multiple interactions at timestamp 0 imply that this user was served a bundle of questions on his first session.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1064189": "While exploring the `train.csv` data, I came across this finding. For 1M rows, the key points I noticed:\n\n- There are 8 such users with multiple interactions at `timestamp == 0`.\n- The common thing among these users are:\n1. They have neighboring content IDs, e.g. 6787, 6788,6786,6785.\n2. They have exactly the same `prior_question_elapsed_time` and `task_container_id`.\n\nCan anyone give any explanation for this occurrence? \nPS: I have explored these findings for `timestamp==0` only. There might be further occurrences, but I couldn't find any explanation for this. Is it that the user is solving those questions or started solving the questions at once? Or the question bundle? But they have different yet neighboring `content_id`. Hence, do `content_id` preserves some sequential information?\nHere is my notebook: https://www.kaggle.com/mrutyunjaybiswal/riiid-aied-challenge-2020-data-exploration-and-fe", "1064196": "This is expected behavior, it just represents a bundle of questions. The `timestamp` indicates when a single task is completed relative to the first task completed, and a single `task_container_id` for a bundle shares its timestamp across all the individual questions that are part of that bundle. `Prior_question_elapsed_time` is also shared across the entire bundle. So in summary, these question bundles are totally normal and occur often in the data, you're just looking at the subset of them that occurred at the beginning of user history (which seems fairly uncommon, given your findings).", "1064304": "Since these are all timestamp =0 interactions, prior_question_elapsed_time must be NA/0 for all. Multiple interactions at timestamp 0 imply that this user was served a bundle of questions on his first session." }, "source": "meta" }