The target binary variable is calculated by observing 18 months performance window after the latest credit card statement, and if the customer does not pay due amount in 120 days after their latest statement date it is considered a default event.
\nLast Statement date is 2019-10-31. So, performance window is from 2019-11-01 to 2021-04-30. This period overlaps with Covid Pandemic Period (Jan 2020 to Aug 2021). Covid pandemic period is a disruption.
\nAre we in for a surprise?
", "messages": [ { "id": "1807420", "postDate": "06/01/2022 02:17:45", "content": "The target binary variable is calculated by observing 18 months performance window after the latest credit card statement, and if the customer does not pay due amount in 120 days after their latest statement date it is considered a default event.
\nLast Statement date is 2019-10-31. So, performance window is from 2019-11-01 to 2021-04-30. This period overlaps with Covid Pandemic Period (Jan 2020 to Aug 2021). Covid pandemic period is a disruption.
\nAre we in for a surprise?
", "rawMarkdown": "*The target binary variable is calculated by observing 18 months performance window after the latest credit card statement, and if the customer does not pay due amount in 120 days after their latest statement date it is considered a default event.*\n\nLast Statement date is 2019-10-31. So, performance window is from 2019-11-01 to 2021-04-30. This period overlaps with Covid Pandemic Period (Jan 2020 to Aug 2021). Covid pandemic period is a disruption.\n\nAre we in for a surprise?", "votes": null }, { "id": "1808139", "postDate": "06/01/2022 14:57:22", "content": "I don't think other data supports a big surprise as a result of Covid. When I have nothing to do (almost always as a retired 75 year old ugly single male) I like to look at trends on \"Google Trends\" It's a great place to get a little bit of domain knowledge for some kaggle challenges.
\nIf you do any type of long term trend (2004-present is the current longest drop down) for any version of terms like \"credit card default\" the trends I saw indicate that the private test period has less credit card default than the period of time for our training data.
\nOur training data was sub sampled at 5% and from the data description you might guess that train and test have the same 5% - if they do and those general trends apply than the test private should have a slightly lower number of defaults.
\nI am going to assume that world wide download of data from google trends can give me the post process adjustment needed.
", "rawMarkdown": "I don't think other data supports a big surprise as a result of Covid. When I have nothing to do (almost always as a retired 75 year old ugly single male) I like to look at trends on [\"Google Trends\" ](https://trends.google.com/trends/explore?date=all&q=credit%20card%20default) It's a great place to get a little bit of domain knowledge for some kaggle challenges.\n\nIf you do any type of long term trend (2004-present is the current longest drop down) for any version of terms like \"credit card default\" the trends I saw indicate that the private test period has less credit card default than the period of time for our training data.\n\nOur training data was sub sampled at 5% and from the data description you might guess that train and test have the same 5% - if they do and those general trends apply than the test private should have a slightly lower number of defaults. \n\nI am going to assume that world wide download of data from google trends can give me the post process adjustment needed.", "votes": null }, { "id": "1808186", "postDate": "06/01/2022 15:36:43", "content": "Thanks for your inputs.
", "rawMarkdown": "Thanks for your inputs.", "votes": null }, { "id": "1808415", "postDate": "06/01/2022 19:33:30", "content": "There are some excellent points there from PC Jimmy. In many places people saved more money during pandemic lockdowns because there was less opportunity to go out and spend money. The savings rate in many places rose to levels not seem for a long time. That would suggest defaults may be lower overall, but the specific details at lower levels may vary based on a number of other factors such as the industry in which the cardholder worked.
", "rawMarkdown": "There are some excellent points there from PC Jimmy. In many places people saved more money during pandemic lockdowns because there was less opportunity to go out and spend money. The savings rate in many places rose to levels not seem for a long time. That would suggest defaults may be lower overall, but the specific details at lower levels may vary based on a number of other factors such as the industry in which the cardholder worked.", "votes": null } ], "comments": [ { "id": 1808139, "author_name": "pcjimmmy", "author_url": "", "post_date": "06/01/2022 14:57:22", "content": "I don't think other data supports a big surprise as a result of Covid. When I have nothing to do (almost always as a retired 75 year old ugly single male) I like to look at trends on \"Google Trends\" It's a great place to get a little bit of domain knowledge for some kaggle challenges.
\nIf you do any type of long term trend (2004-present is the current longest drop down) for any version of terms like \"credit card default\" the trends I saw indicate that the private test period has less credit card default than the period of time for our training data.
\nOur training data was sub sampled at 5% and from the data description you might guess that train and test have the same 5% - if they do and those general trends apply than the test private should have a slightly lower number of defaults.
\nI am going to assume that world wide download of data from google trends can give me the post process adjustment needed.
", "votes": null, "replies": [ { "id": 1808186, "author_name": "balajisr", "author_url": "", "post_date": "06/01/2022 15:36:43", "content": "Thanks for your inputs.
", "votes": null, "replies": [] }, { "id": 1808415, "author_name": "datahobbit", "author_url": "", "post_date": "06/01/2022 19:33:30", "content": "There are some excellent points there from PC Jimmy. In many places people saved more money during pandemic lockdowns because there was less opportunity to go out and spend money. The savings rate in many places rose to levels not seem for a long time. That would suggest defaults may be lower overall, but the specific details at lower levels may vary based on a number of other factors such as the industry in which the cardholder worked.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1807420": "*The target binary variable is calculated by observing 18 months performance window after the latest credit card statement, and if the customer does not pay due amount in 120 days after their latest statement date it is considered a default event.*\n\nLast Statement date is 2019-10-31. So, performance window is from 2019-11-01 to 2021-04-30. This period overlaps with Covid Pandemic Period (Jan 2020 to Aug 2021). Covid pandemic period is a disruption.\n\nAre we in for a surprise?", "1808139": "I don't think other data supports a big surprise as a result of Covid. When I have nothing to do (almost always as a retired 75 year old ugly single male) I like to look at trends on [\"Google Trends\" ](https://trends.google.com/trends/explore?date=all&q=credit%20card%20default) It's a great place to get a little bit of domain knowledge for some kaggle challenges.\n\nIf you do any type of long term trend (2004-present is the current longest drop down) for any version of terms like \"credit card default\" the trends I saw indicate that the private test period has less credit card default than the period of time for our training data.\n\nOur training data was sub sampled at 5% and from the data description you might guess that train and test have the same 5% - if they do and those general trends apply than the test private should have a slightly lower number of defaults. \n\nI am going to assume that world wide download of data from google trends can give me the post process adjustment needed.", "1808186": "Thanks for your inputs.", "1808415": "There are some excellent points there from PC Jimmy. In many places people saved more money during pandemic lockdowns because there was less opportunity to go out and spend money. The savings rate in many places rose to levels not seem for a long time. That would suggest defaults may be lower overall, but the specific details at lower levels may vary based on a number of other factors such as the industry in which the cardholder worked." }, "source": "meta" }