Sampling Rate in Bootstrap Forest

Question

This, There are big size data.

Response(Y) is categorical data, good or bad. Total numbers of rows are over 10 million.

But, most of them are 'good' data on the other hand the numbers of 'bad' data are just twenty, or something. Because the numbers of 'bad' data is too small, we failed in analyzing this data, especially in case of Bootstrap Forest.

Please let me know good solution how we settle such a ultimate unbalanced case sregardless of bootstrap forest or other things.

Thanks

IkjuShin · Answer

Hi Martin,

Very Good Reference !

Thanks !

martindemel · Answer

Michael Crotty has held a presentation at Discovery Europe about different strategies to handle such issues. You may find something useful in it:

https://community.jmp.com/t5/Discovery-Summit-Europe-2019/Anomaly-Detection-and-JMP-R-Pro-2019-EU-45MP-115/ta-p/110065

Martin