{ "id": 43387, "title": "how come this user is churn? SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg=", "url": "/competitions/kkbox-churn-prediction-challenge/discussion/43387", "author_name": "", "post_date": "2017-11-14T03:58:35.370100500Z", "votes": null, "comment_count": 9, "views": 0, "content": "

transactions[transactions['msno']=='SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg='].head(10)\nOut[102]: \n msno payment_method_id \\\n14793958 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38
\n16290331 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 28
\n17042391 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38
\n20984852 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 41

\n\n
      payment_plan_days  plan_list_price  actual_amount_paid  \\\n
\n\n

14793958 360 1200 1200
\n16290331 60 300 300
\n17042391 360 1200 1200
\n20984852 30 149 149

\n\n
      is_auto_renew  transaction_date  membership_expire_date  is_cancel  \n
\n\n

14793958 0 20160130 20170124 0
\n16290331 0 20160715 20170325 0
\n17042391 0 20150101 20151227 0
\n20984852 1 20150101 20141231 1

", "messages": [ { "id": "243451", "postDate": "11/14/2017 03:58:35", "content": "

transactions[transactions['msno']=='SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg='].head(10)\nOut[102]: \n msno payment_method_id \\\n14793958 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38
\n16290331 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 28
\n17042391 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38
\n20984852 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 41

\n\n
      payment_plan_days  plan_list_price  actual_amount_paid  \\\n
\n\n

14793958 360 1200 1200
\n16290331 60 300 300
\n17042391 360 1200 1200
\n20984852 30 149 149

\n\n
      is_auto_renew  transaction_date  membership_expire_date  is_cancel  \n
\n\n

14793958 0 20160130 20170124 0
\n16290331 0 20160715 20170325 0
\n17042391 0 20150101 20151227 0
\n20984852 1 20150101 20141231 1

", "rawMarkdown": "transactions[transactions['msno']=='SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg='].head(10)\nOut[102]: \n msno payment_method_id \\\n14793958 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38 \n16290331 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 28 \n17042391 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38 \n20984852 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 41 \n\n payment_plan_days plan_list_price actual_amount_paid \\\n14793958 360 1200 1200 \n16290331 60 300 300 \n17042391 360 1200 1200 \n20984852 30 149 149 \n\n is_auto_renew transaction_date membership_expire_date is_cancel \n14793958 0 20160130 20170124 0 \n16290331 0 20160715 20170325 0 \n17042391 0 20150101 20151227 0 \n20984852 1 20150101 20141231 1", "votes": null }, { "id": "243454", "postDate": "11/14/2017 04:08:06", "content": "

also this one: Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk=

\n\n

never cancelled, how come 'is_churn'?

\n\n

transactions[transactions['msno']=='Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk='].head(100)

\n\n
                                              msno  payment_method_id  \\\n
\n\n

94652 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n863296 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n1786733 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n3859014 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n6179436 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n6412330 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n7695458 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n7720839 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n8172018 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n15279713 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n15370824 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n16191725 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n16605762 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n20650995 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36

\n\n
      payment_plan_days  plan_list_price  actual_amount_paid  \\\n
\n\n

94652 30 180 180
\n863296 30 150 150
\n1786733 30 150 150
\n3859014 30 180 180
\n6179436 30 180 180
\n6412330 30 180 180
\n7695458 30 180 180
\n7720839 30 150 150
\n8172018 30 150 150
\n15279713 30 150 150
\n15370824 30 180 180
\n16191725 30 180 180
\n16605762 30 180 180
\n20650995 30 180 180

\n\n
      is_auto_renew  transaction_date  membership_expire_date  is_cancel  \n
\n\n

94652 0 20161004 20161103 0
\n863296 0 20160124 20160223 0
\n1786733 0 20160327 20160426 0
\n3859014 0 20160427 20160527 0
\n6179436 0 20160801 20160831 0
\n6412330 0 20160901 20161001 0
\n7695458 0 20161122 20161222 0
\n7720839 0 20151116 20151216 0
\n8172018 0 20151218 20160117 0
\n15279713 0 20160225 20160326 0
\n15370824 0 20160628 20160728 0
\n16191725 0 20160528 20160627 0
\n16605762 0 20170209 20170311 0
\n20650995 0 20161224 20170123 0

", "rawMarkdown": "also this one: Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= \n\nnever cancelled, how come 'is_churn'?\n\ntransactions[transactions['msno']=='Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk='].head(100)\n\n msno payment_method_id \\\n94652 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n863296 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n1786733 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n3859014 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n6179436 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n6412330 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n7695458 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n7720839 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n8172018 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n15279713 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n15370824 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n16191725 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n16605762 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n20650995 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n\n payment_plan_days plan_list_price actual_amount_paid \\\n94652 30 180 180 \n863296 30 150 150 \n1786733 30 150 150 \n3859014 30 180 180 \n6179436 30 180 180 \n6412330 30 180 180 \n7695458 30 180 180 \n7720839 30 150 150 \n8172018 30 150 150 \n15279713 30 150 150 \n15370824 30 180 180 \n16191725 30 180 180 \n16605762 30 180 180 \n20650995 30 180 180 \n\n is_auto_renew transaction_date membership_expire_date is_cancel \n94652 0 20161004 20161103 0 \n863296 0 20160124 20160223 0 \n1786733 0 20160327 20160426 0 \n3859014 0 20160427 20160527 0 \n6179436 0 20160801 20160831 0 \n6412330 0 20160901 20161001 0 \n7695458 0 20161122 20161222 0 \n7720839 0 20151116 20151216 0 \n8172018 0 20151218 20160117 0 \n15279713 0 20160225 20160326 0 \n15370824 0 20160628 20160728 0 \n16191725 0 20160528 20160627 0 \n16605762 0 20170209 20170311 0 \n20650995 0 20161224 20170123 0", "votes": null }, { "id": "243545", "postDate": "11/14/2017 09:57:02", "content": "

With introduction of the new transaction logs, the test set may contain entries that change test user labels.\nWe have provided the scala/spark code to generate test and training labels. You can use it or modify it to generate labels for the training set.

", "rawMarkdown": "With introduction of the new transaction logs, the test set may contain entries that change test user labels.\nWe have provided the scala/spark code to generate test and training labels. You can use it or modify it to generate labels for the training set.", "votes": null }, { "id": "243875", "postDate": "11/15/2017 03:09:31", "content": "

does any one have a good explanation how come these two users are 'is_churn' users? they look to be good standing users till mid of march of 2017

", "rawMarkdown": "does any one have a good explanation how come these two users are 'is_churn' users? they look to be good standing users till mid of march of 2017", "votes": null }, { "id": "246344", "postDate": "11/21/2017 01:17:51", "content": "

I think churning and canceling are different. It seems a user can churn without canceling and cancel without churning. The administrators define churning as a user not setting up any new membership within 30 days of their current membership expiring, regardless of whether they cancel or not. So you might see a user cancel but then sign up again before 30 days is up and then not labeled churn. Or you might see a user not cancel, but also not resubscribe within 30 days and then they are labeled churn. Does this help at all?

", "rawMarkdown": "I think churning and canceling are different. It seems a user can churn without canceling and cancel without churning. The administrators define churning as a user not setting up any new membership within 30 days of their current membership expiring, regardless of whether they cancel or not. So you might see a user cancel but then sign up again before 30 days is up and then not labeled churn. Or you might see a user not cancel, but also not resubscribe within 30 days and then they are labeled churn. Does this help at all?", "votes": null }, { "id": "246577", "postDate": "11/21/2017 11:02:19", "content": "

Thanks for your remark butfor me it remains inconsistency in the link between churn/transaction_date and membership_expiration_date.... Using as example user : \n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222\nIn transactions_logs, how can such user can have a membership_expire in 2017 with last transaction in december 2015? In particular since there is no plan for more than 30 days\nCorresponding line in transactions.csv :

\n\n

//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222, 20170211 ,0

\n\n

It becomes difficult to predict membership_expiration based just on data in transactions for me. Am I missing something?

", "rawMarkdown": "Thanks for your remark butfor me it remains inconsistency in the link between churn/transaction_date and membership_expiration_date.... Using as example user : \n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222\nIn transactions_logs, how can such user can have a membership_expire in 2017 with last transaction in december 2015? In particular since there is no plan for more than 30 days\nCorresponding line in transactions.csv :\n\n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,**20151222**, **20170211** ,0\n\nIt becomes difficult to predict membership_expiration based just on data in transactions for me. Am I missing something?", "votes": null }, { "id": "246995", "postDate": "11/22/2017 04:35:58", "content": "

The user you posted at the top churned in January of 2016, according to the output you provided. Their previous plan ended Dec 27 2015 and they didn't sign up again until Jan 31 2016, which qualifies for user churn by admin's definition. The second user you posted underneath that does not seem to have churned.

", "rawMarkdown": "The user you posted at the top churned in January of 2016, according to the output you provided. Their previous plan ended Dec 27 2015 and they didn't sign up again until Jan 31 2016, which qualifies for user churn by admin's definition. The second user you posted underneath that does not seem to have churned.", "votes": null }, { "id": "246996", "postDate": "11/22/2017 04:38:43", "content": "

Yeah it is weird. Based on the output provided I cannot troubleshoot. But I agree this dataset is incredibly confusing : )

", "rawMarkdown": "Yeah it is weird. Based on the output provided I cannot troubleshoot. But I agree this dataset is incredibly confusing : )", "votes": null }, { "id": "247003", "postDate": "11/22/2017 04:57:52", "content": "

how about 2nd example?

\n\n

I think we predict churn in the coming month. it is easy if just figure out churn in past data.

", "rawMarkdown": "how about 2nd example?\n\nI think we predict churn in the coming month. it is easy if just figure out churn in past data.", "votes": null }, { "id": "247012", "postDate": "11/22/2017 05:30:41", "content": "

I agree

", "rawMarkdown": "I agree", "votes": null } ], "comments": [ { "id": 243454, "author_name": "yliu9999", "author_url": "", "post_date": "11/14/2017 04:08:06", "content": "

also this one: Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk=

\n\n

never cancelled, how come 'is_churn'?

\n\n

transactions[transactions['msno']=='Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk='].head(100)

\n\n
                                              msno  payment_method_id  \\\n
\n\n

94652 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n863296 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n1786733 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n3859014 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n6179436 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n6412330 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n7695458 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n7720839 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n8172018 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n15279713 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n15370824 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n16191725 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n16605762 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36
\n20650995 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36

\n\n
      payment_plan_days  plan_list_price  actual_amount_paid  \\\n
\n\n

94652 30 180 180
\n863296 30 150 150
\n1786733 30 150 150
\n3859014 30 180 180
\n6179436 30 180 180
\n6412330 30 180 180
\n7695458 30 180 180
\n7720839 30 150 150
\n8172018 30 150 150
\n15279713 30 150 150
\n15370824 30 180 180
\n16191725 30 180 180
\n16605762 30 180 180
\n20650995 30 180 180

\n\n
      is_auto_renew  transaction_date  membership_expire_date  is_cancel  \n
\n\n

94652 0 20161004 20161103 0
\n863296 0 20160124 20160223 0
\n1786733 0 20160327 20160426 0
\n3859014 0 20160427 20160527 0
\n6179436 0 20160801 20160831 0
\n6412330 0 20160901 20161001 0
\n7695458 0 20161122 20161222 0
\n7720839 0 20151116 20151216 0
\n8172018 0 20151218 20160117 0
\n15279713 0 20160225 20160326 0
\n15370824 0 20160628 20160728 0
\n16191725 0 20160528 20160627 0
\n16605762 0 20170209 20170311 0
\n20650995 0 20161224 20170123 0

", "votes": null, "replies": [] }, { "id": 243545, "author_name": "ardenkkbox", "author_url": "", "post_date": "11/14/2017 09:57:02", "content": "

With introduction of the new transaction logs, the test set may contain entries that change test user labels.\nWe have provided the scala/spark code to generate test and training labels. You can use it or modify it to generate labels for the training set.

", "votes": null, "replies": [] }, { "id": 243875, "author_name": "yliu9999", "author_url": "", "post_date": "11/15/2017 03:09:31", "content": "

does any one have a good explanation how come these two users are 'is_churn' users? they look to be good standing users till mid of march of 2017

", "votes": null, "replies": [ { "id": 246995, "author_name": "davischumacher", "author_url": "", "post_date": "11/22/2017 04:35:58", "content": "

The user you posted at the top churned in January of 2016, according to the output you provided. Their previous plan ended Dec 27 2015 and they didn't sign up again until Jan 31 2016, which qualifies for user churn by admin's definition. The second user you posted underneath that does not seem to have churned.

", "votes": null, "replies": [] }, { "id": 247003, "author_name": "yliu9999", "author_url": "", "post_date": "11/22/2017 04:57:52", "content": "

how about 2nd example?

\n\n

I think we predict churn in the coming month. it is easy if just figure out churn in past data.

", "votes": null, "replies": [] }, { "id": 247012, "author_name": "davischumacher", "author_url": "", "post_date": "11/22/2017 05:30:41", "content": "

I agree

", "votes": null, "replies": [] } ] }, { "id": 246344, "author_name": "davischumacher", "author_url": "", "post_date": "11/21/2017 01:17:51", "content": "

I think churning and canceling are different. It seems a user can churn without canceling and cancel without churning. The administrators define churning as a user not setting up any new membership within 30 days of their current membership expiring, regardless of whether they cancel or not. So you might see a user cancel but then sign up again before 30 days is up and then not labeled churn. Or you might see a user not cancel, but also not resubscribe within 30 days and then they are labeled churn. Does this help at all?

", "votes": null, "replies": [ { "id": 246577, "author_name": "guiyom", "author_url": "", "post_date": "11/21/2017 11:02:19", "content": "

Thanks for your remark butfor me it remains inconsistency in the link between churn/transaction_date and membership_expiration_date.... Using as example user : \n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222\nIn transactions_logs, how can such user can have a membership_expire in 2017 with last transaction in december 2015? In particular since there is no plan for more than 30 days\nCorresponding line in transactions.csv :

\n\n

//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222, 20170211 ,0

\n\n

It becomes difficult to predict membership_expiration based just on data in transactions for me. Am I missing something?

", "votes": null, "replies": [] }, { "id": 246996, "author_name": "davischumacher", "author_url": "", "post_date": "11/22/2017 04:38:43", "content": "

Yeah it is weird. Based on the output provided I cannot troubleshoot. But I agree this dataset is incredibly confusing : )

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "243451": "transactions[transactions['msno']=='SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg='].head(10)\nOut[102]: \n msno payment_method_id \\\n14793958 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38 \n16290331 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 28 \n17042391 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 38 \n20984852 SlkPQrh6oIqsdpW+Sl2adYlJNEwWSQsIYycM11dSUQg= 41 \n\n payment_plan_days plan_list_price actual_amount_paid \\\n14793958 360 1200 1200 \n16290331 60 300 300 \n17042391 360 1200 1200 \n20984852 30 149 149 \n\n is_auto_renew transaction_date membership_expire_date is_cancel \n14793958 0 20160130 20170124 0 \n16290331 0 20160715 20170325 0 \n17042391 0 20150101 20151227 0 \n20984852 1 20150101 20141231 1", "243454": "also this one: Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= \n\nnever cancelled, how come 'is_churn'?\n\ntransactions[transactions['msno']=='Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk='].head(100)\n\n msno payment_method_id \\\n94652 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n863296 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n1786733 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n3859014 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n6179436 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n6412330 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n7695458 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n7720839 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n8172018 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n15279713 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n15370824 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n16191725 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n16605762 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n20650995 Q/S7jZ/6XtHOaJStQaM89sdC8e5yn22KwkR4PzgJzFk= 36 \n\n payment_plan_days plan_list_price actual_amount_paid \\\n94652 30 180 180 \n863296 30 150 150 \n1786733 30 150 150 \n3859014 30 180 180 \n6179436 30 180 180 \n6412330 30 180 180 \n7695458 30 180 180 \n7720839 30 150 150 \n8172018 30 150 150 \n15279713 30 150 150 \n15370824 30 180 180 \n16191725 30 180 180 \n16605762 30 180 180 \n20650995 30 180 180 \n\n is_auto_renew transaction_date membership_expire_date is_cancel \n94652 0 20161004 20161103 0 \n863296 0 20160124 20160223 0 \n1786733 0 20160327 20160426 0 \n3859014 0 20160427 20160527 0 \n6179436 0 20160801 20160831 0 \n6412330 0 20160901 20161001 0 \n7695458 0 20161122 20161222 0 \n7720839 0 20151116 20151216 0 \n8172018 0 20151218 20160117 0 \n15279713 0 20160225 20160326 0 \n15370824 0 20160628 20160728 0 \n16191725 0 20160528 20160627 0 \n16605762 0 20170209 20170311 0 \n20650995 0 20161224 20170123 0", "243545": "With introduction of the new transaction logs, the test set may contain entries that change test user labels.\nWe have provided the scala/spark code to generate test and training labels. You can use it or modify it to generate labels for the training set.", "243875": "does any one have a good explanation how come these two users are 'is_churn' users? they look to be good standing users till mid of march of 2017", "246344": "I think churning and canceling are different. It seems a user can churn without canceling and cancel without churning. The administrators define churning as a user not setting up any new membership within 30 days of their current membership expiring, regardless of whether they cancel or not. So you might see a user cancel but then sign up again before 30 days is up and then not labeled churn. Or you might see a user not cancel, but also not resubscribe within 30 days and then they are labeled churn. Does this help at all?", "246577": "Thanks for your remark butfor me it remains inconsistency in the link between churn/transaction_date and membership_expiration_date.... Using as example user : \n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,20151222\nIn transactions_logs, how can such user can have a membership_expire in 2017 with last transaction in december 2015? In particular since there is no plan for more than 30 days\nCorresponding line in transactions.csv :\n\n//7e1ZBmuIEL9oT0OxqUcI5NavDuxraEtx5+jiNPwxI=,41,30,149,149,1,**20151222**, **20170211** ,0\n\nIt becomes difficult to predict membership_expiration based just on data in transactions for me. Am I missing something?", "246995": "The user you posted at the top churned in January of 2016, according to the output you provided. Their previous plan ended Dec 27 2015 and they didn't sign up again until Jan 31 2016, which qualifies for user churn by admin's definition. The second user you posted underneath that does not seem to have churned.", "246996": "Yeah it is weird. Based on the output provided I cannot troubleshoot. But I agree this dataset is incredibly confusing : )", "247003": "how about 2nd example?\n\nI think we predict churn in the coming month. it is easy if just figure out churn in past data.", "247012": "I agree" }, "source": "meta" }