I am a bit confused by the following example given in the explanation section for data not falling within the scope for prediction\n(2017-01-01, 2017-02-28, false)
\n\n(2017-02-25, 2017-04-03, false)
\n\n(2017-03-15, 2017-03-16, true)
\n\n(2017-03-18, 2017-04-02, false)
\n\nHere the instruction says the membership expiration date is 4/2 which is outside the month of March thus this data is out of scope. But if this is true, it will mean we only need to predict the users with expiration date in March and then renew within 30 days but have to renew in April. Is that right? Otherwise if that user renew again in March with an expiration beyond March will never be predicted, but in reality it looks to me they should be treated as Churn=0 since they did renew. Am I missing something here?
", "messages": [ { "id": "225469", "postDate": "09/29/2017 06:15:40", "content": "I am a bit confused by the following example given in the explanation section for data not falling within the scope for prediction\n(2017-01-01, 2017-02-28, false)
\n\n(2017-02-25, 2017-04-03, false)
\n\n(2017-03-15, 2017-03-16, true)
\n\n(2017-03-18, 2017-04-02, false)
\n\nHere the instruction says the membership expiration date is 4/2 which is outside the month of March thus this data is out of scope. But if this is true, it will mean we only need to predict the users with expiration date in March and then renew within 30 days but have to renew in April. Is that right? Otherwise if that user renew again in March with an expiration beyond March will never be predicted, but in reality it looks to me they should be treated as Churn=0 since they did renew. Am I missing something here?
", "rawMarkdown": "I am a bit confused by the following example given in the explanation section for data not falling within the scope for prediction\n(2017-01-01, 2017-02-28, false)\n\n(2017-02-25, 2017-04-03, false)\n\n(2017-03-15, 2017-03-16, true)\n\n(2017-03-18, 2017-04-02, false)\n\nHere the instruction says the membership expiration date is 4/2 which is outside the month of March thus this data is out of scope. But if this is true, it will mean we only need to predict the users with expiration date in March and then renew within 30 days but have to renew in April. Is that right? Otherwise if that user renew again in March with an expiration beyond March will never be predicted, but in reality it looks to me they should be treated as Churn=0 since they did renew. Am I missing something here?", "votes": null }, { "id": "225470", "postDate": "09/29/2017 06:20:27", "content": "Adding to my question, the definition of 'out of scope' will mean none of the auto-renew user with 30 day subscription period will be predicted. The reason being if the expiration date falls on a date in March, it will auto renew the next day and then result in an expiration date in April, is that the true and the intention of this competition?
", "rawMarkdown": "Adding to my question, the definition of 'out of scope' will mean none of the auto-renew user with 30 day subscription period will be predicted. The reason being if the expiration date falls on a date in March, it will auto renew the next day and then result in an expiration date in April, is that the true and the intention of this competition?", "votes": null }, { "id": "226361", "postDate": "10/02/2017 02:14:45", "content": "Yes, the task is to \"only predict the users with expiration date in March and then renew within 30 days but have to renew in April.\" The 30 days threshold is selected for this competition. In reality, it can be adjusted to meet the business requirement.
", "rawMarkdown": "Yes, the task is to \"only predict the users with expiration date in March and then renew within 30 days but have to renew in April.\" The 30 days threshold is selected for this competition. In reality, it can be adjusted to meet the business requirement.", "votes": null }, { "id": "226732", "postDate": "10/03/2017 00:59:57", "content": "Thanks for the answer. Now I seem to understand you are trying to predict only subscribers who 'paused' their subscriptions in March and whether or not they are coming back in April. That's a very interesting proposition. Maybe that's a very big population, I think I will find out through the data...
", "rawMarkdown": "Thanks for the answer. Now I seem to understand you are trying to predict only subscribers who 'paused' their subscriptions in March and whether or not they are coming back in April. That's a very interesting proposition. Maybe that's a very big population, I think I will find out through the data...", "votes": null }, { "id": "226778", "postDate": "10/03/2017 03:28:02", "content": "the word paused is not exactly true. The user the didn't renew their subscription within 30 day of member expiration include those paused and those left forever (well, or up to this moment)
", "rawMarkdown": "the word paused is not exactly true. The user the didn't renew their subscription within 30 day of member expiration include those paused and those left forever (well, or up to this moment)", "votes": null }, { "id": "228207", "postDate": "10/06/2017 04:55:56", "content": "Hi, given the above definition of 'out of scope' for the prediction set, I am wondering why would this user belong to the training set:\n msno is_churn is_auto_renew \\\n0 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 0
\n1 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 1
transaction_date membership_expire_date is_cancel
\n0 2017-02-15 2017-03-17 0
\n1 2017-02-15 2017-02-13 1
This user has a membership_expire_date in March at 3/17, it's outside 02/17, so he/she should not be in the train_set, is that right?
", "rawMarkdown": "Hi, given the above definition of 'out of scope' for the prediction set, I am wondering why would this user belong to the training set:\n msno is_churn is_auto_renew \\\n0 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 0 \n1 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 1 \n\n transaction_date membership_expire_date is_cancel \n0 2017-02-15 2017-03-17 0 \n1 2017-02-15 2017-02-13 1 \n\nThis user has a membership_expire_date in March at 3/17, it's outside 02/17, so he/she should not be in the train_set, is that right?", "votes": null }, { "id": "228237", "postDate": "10/06/2017 07:16:58", "content": "Since you gave example of the user out of scope, does that mean in your prediction set, there are users that do not need to be predicted? In other words, for some users in prediction set, it does not matter I predict a 0 or 1? But we will not really know who those users are?
", "rawMarkdown": "Since you gave example of the user out of scope, does that mean in your prediction set, there are users that do not need to be predicted? In other words, for some users in prediction set, it does not matter I predict a 0 or 1? But we will not really know who those users are?", "votes": null }, { "id": "228244", "postDate": "10/06/2017 07:34:02", "content": "And why would this user be considered a 'churned' user?\n msno is_churn city bd gender \\\n134 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n143 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n145 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n128 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
registered_via registration_init_time expiration_date payment_method_id \\\n134 9 2006-12-22 2017-09-21 38
\n143 9 2006-12-22 2017-09-21 38
\n145 9 2006-12-22 2017-09-21 38
\n128 9 2006-12-22 2017-09-21 38
payment_plan_days plan_list_price actual_amount_paid is_auto_renew \\\n134 30 149 149 0
\n143 30 149 149 0
\n145 30 149 149 0
\n128 30 149 149 0
transaction_date membership_expire_date is_cancel \n134 2016-11-21 2016-12-21 0
\n143 2016-12-16 2017-01-20 0
\n145 2017-01-17 2017-02-19 0
\n128 2017-02-16 2017-03-21 0
Adding to my question, the definition of 'out of scope' will mean none of the auto-renew user with 30 day subscription period will be predicted. The reason being if the expiration date falls on a date in March, it will auto renew the next day and then result in an expiration date in April, is that the true and the intention of this competition?
", "votes": null, "replies": [] }, { "id": 226361, "author_name": "ardenkkbox", "author_url": "", "post_date": "10/02/2017 02:14:45", "content": "Yes, the task is to \"only predict the users with expiration date in March and then renew within 30 days but have to renew in April.\" The 30 days threshold is selected for this competition. In reality, it can be adjusted to meet the business requirement.
", "votes": null, "replies": [] }, { "id": 226732, "author_name": "shunjiangxu", "author_url": "", "post_date": "10/03/2017 00:59:57", "content": "Thanks for the answer. Now I seem to understand you are trying to predict only subscribers who 'paused' their subscriptions in March and whether or not they are coming back in April. That's a very interesting proposition. Maybe that's a very big population, I think I will find out through the data...
", "votes": null, "replies": [ { "id": 226778, "author_name": "ardenkkbox", "author_url": "", "post_date": "10/03/2017 03:28:02", "content": "the word paused is not exactly true. The user the didn't renew their subscription within 30 day of member expiration include those paused and those left forever (well, or up to this moment)
", "votes": null, "replies": [] } ] }, { "id": 228207, "author_name": "shunjiangxu", "author_url": "", "post_date": "10/06/2017 04:55:56", "content": "Hi, given the above definition of 'out of scope' for the prediction set, I am wondering why would this user belong to the training set:\n msno is_churn is_auto_renew \\\n0 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 0
\n1 ++wDR2uEVD8Ep2jLNXD6PDDMbmQZXfL97TJiSqMsvH0= 0 1
transaction_date membership_expire_date is_cancel
\n0 2017-02-15 2017-03-17 0
\n1 2017-02-15 2017-02-13 1
This user has a membership_expire_date in March at 3/17, it's outside 02/17, so he/she should not be in the train_set, is that right?
", "votes": null, "replies": [] }, { "id": 228237, "author_name": "shunjiangxu", "author_url": "", "post_date": "10/06/2017 07:16:58", "content": "Since you gave example of the user out of scope, does that mean in your prediction set, there are users that do not need to be predicted? In other words, for some users in prediction set, it does not matter I predict a 0 or 1? But we will not really know who those users are?
", "votes": null, "replies": [] }, { "id": 228244, "author_name": "shunjiangxu", "author_url": "", "post_date": "10/06/2017 07:34:02", "content": "And why would this user be considered a 'churned' user?\n msno is_churn city bd gender \\\n134 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n143 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n145 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
\n128 I8dFN2EjFN1mt4Xel8WQX1/g7u6Dg4PBMHLkiDjhUS8= 1 4 42 1
registered_via registration_init_time expiration_date payment_method_id \\\n134 9 2006-12-22 2017-09-21 38
\n143 9 2006-12-22 2017-09-21 38
\n145 9 2006-12-22 2017-09-21 38
\n128 9 2006-12-22 2017-09-21 38
payment_plan_days plan_list_price actual_amount_paid is_auto_renew \\\n134 30 149 149 0
\n143 30 149 149 0
\n145 30 149 149 0
\n128 30 149 149 0
transaction_date membership_expire_date is_cancel \n134 2016-11-21 2016-12-21 0
\n143 2016-12-16 2017-01-20 0
\n145 2017-01-17 2017-02-19 0
\n128 2017-02-16 2017-03-21 0