container_id, and this discussion points out the problem on inconsistency between timestamp and task_container_id, as shown in below picture.\n
![\"Figure](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2293617%2Fd37d94623f3bf00934f2479af90f8040%2Fkaggle_riiid_1.png?generation=1610252567534467&alt=media\")
There has been a lot of discussion on container_id, and this discussion points out the problem on inconsistency between timestamp and task_container_id, as shown in below picture.
\n
According to answer by Kaggle staff in other discussion, the timestamp column shows when an activity is finished, not when it started. The task_container_id comes from a scenario where the user gets a batch of questions each time they do a task, and the task_container_id represents the number of batch starting from 0 and increasing, so the task_container_id reflects the order in which the user starts the task. If users work on one batch of questions and then the next batch of questions, then the task_container_id should increase with timestamp. Below picture shows this scenario where one batch of questions is completed and then the next batch is completed (the number in the box represents task_container_id).
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2293617%2F084ad07947075ae2e35f0bb814dd1a1e%2Fkaggle_riiid_2.png?generation=1610253099661857&alt=media\")
But when we were in school, our teachers should have taught us that if we don't know how to answer a question , we should skip it first and do it later. This is the reason for the discontinuous of task_container_id. For example, in below picture, the user doesn't know how to do the questions with task_container_id of 1, so answer the questions with task_container_id of 2 first. Questions with task_container_id of 2 are easy to answer, and then go back to the questions with task_container_id of 1. Thus, the task_container_id does not automatically increment with timestamp. From this phenomenon we can construct a strong feature: the difference between the task_container_id of the current batch and that of the previous batch. The greater the absolute value of the difference, the more the user skip in current question.
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F2293617%2Fcc3c2fa6d6f24b0d62abd53aeedb90f7%2Fkaggle_riiid_3.png?generation=1610253573250499&alt=media\")
This feature increases by about 1.5k.
", "messages": [ { "id": 1146815, "postDate": "2021-01-10T04:42:24.730Z", "content": "There has been a lot of discussion on container_id, and this discussion points out the problem on inconsistency between timestamp and task_container_id, as shown in below picture.
\n
According to answer by Kaggle staff in other discussion, the timestamp column shows when an activity is finished, not when it started. The task_container_id comes from a scenario where the user gets a batch of questions each time they do a task, and the task_container_id represents the number of batch starting from 0 and increasing, so the task_container_id reflects the order in which the user starts the task. If users work on one batch of questions and then the next batch of questions, then the task_container_id should increase with timestamp. Below picture shows this scenario where one batch of questions is completed and then the next batch is completed (the number in the box represents task_container_id).
But when we were in school, our teachers should have taught us that if we don't know how to answer a question , we should skip it first and do it later. This is the reason for the discontinuous of task_container_id. For example, in below picture, the user doesn't know how to do the questions with task_container_id of 1, so answer the questions with task_container_id of 2 first. Questions with task_container_id of 2 are easy to answer, and then go back to the questions with task_container_id of 1. Thus, the task_container_id does not automatically increment with timestamp. From this phenomenon we can construct a strong feature: the difference between the task_container_id of the current batch and that of the previous batch. The greater the absolute value of the difference, the more the user skip in current question.
This feature increases by about 1.5k.
", "rawMarkdown": "There has been a lot of discussion on `container_id`, and [this discussion](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465) points out the problem on inconsistency between `timestamp` and `task_container_id`, as shown in below picture.\n\n\nAccording to answer by Kaggle staff in other [discussion](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189351#1041405), the `timestamp` column shows when an activity is finished, not when it started. The `task_container_id` comes from a scenario where the user gets a batch of questions each time they do a task, and the `task_container_id` represents the number of batch starting from 0 and increasing, so the `task_container_id` reflects the order in which the user starts the task. If users work on one batch of questions and then the next batch of questions, then the `task_container_id` should increase with `timestamp`. Below picture shows this scenario where one batch of questions is completed and then the next batch is completed (the number in the box represents `task_container_id`).\n\n\n\nBut when we were in school, our teachers should have taught us that if we don't know how to answer a question , we should skip it first and do it later. This is the reason for the discontinuous of `task_container_id`. For example, in below picture, the user doesn't know how to do the questions with `task_container_id` of 1, so answer the questions with `task_container_id` of 2 first. Questions with `task_container_id` of 2 are easy to answer, and then go back to the questions with `task_container_id` of 1. Thus, the `task_container_id` does not automatically increment with `timestamp`. From this phenomenon we can construct a strong feature: the difference between the `task_container_id` of the current batch and that of the previous batch. The greater the absolute value of the difference, the more the user skip in current question. \n\n\n\n**This feature increases by about 1.5k.**", "votes": 8 } ], "comments": [], "raw_markdown_by_id": { "1146815": "There has been a lot of discussion on `container_id`, and [this discussion](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189465) points out the problem on inconsistency between `timestamp` and `task_container_id`, as shown in below picture.\n\n\nAccording to answer by Kaggle staff in other [discussion](https://www.kaggle.com/c/riiid-test-answer-prediction/discussion/189351#1041405), the `timestamp` column shows when an activity is finished, not when it started. The `task_container_id` comes from a scenario where the user gets a batch of questions each time they do a task, and the `task_container_id` represents the number of batch starting from 0 and increasing, so the `task_container_id` reflects the order in which the user starts the task. If users work on one batch of questions and then the next batch of questions, then the `task_container_id` should increase with `timestamp`. Below picture shows this scenario where one batch of questions is completed and then the next batch is completed (the number in the box represents `task_container_id`).\n\n\n\nBut when we were in school, our teachers should have taught us that if we don't know how to answer a question , we should skip it first and do it later. This is the reason for the discontinuous of `task_container_id`. For example, in below picture, the user doesn't know how to do the questions with `task_container_id` of 1, so answer the questions with `task_container_id` of 2 first. Questions with `task_container_id` of 2 are easy to answer, and then go back to the questions with `task_container_id` of 1. Thus, the `task_container_id` does not automatically increment with `timestamp`. From this phenomenon we can construct a strong feature: the difference between the `task_container_id` of the current batch and that of the previous batch. The greater the absolute value of the difference, the more the user skip in current question. \n\n\n\n**This feature increases by about 1.5k.**" } }