How can I apply SMOTE for multi-class Classification of NSL-KDD data set in RapidMiner?

YeshSWJTU

I am working on feature selection in  Network Intrusion Detection System (NIDS) using  NSL-KDD
data set.
How can I apply SMOTE for multi-class Classification of NSL-KDD data set in RapidMiner?

In NSL-KDD data set, there are five classes Normal, DoS, Probe, U2R and R2L. But the classes are extremely imbalanced specially U2R and R2L. I am trying to balance this data set using SMOTE and to dynamically balance the data set. But i am getting problem to solve using rapid miner. I can apply SMOTE using WEKA but, i need to balance dynamically using RapidMiner. I need your help Thank you

Telcontar120

There is a SMOTE Upsampling operator for RapidMiner in the Operator Toolbox, which you will need to install from the extension marketplace (it is free).  You might also consider using weighting instead of upsampling depending on the ML algorithm you intend to use.  There are several suitable weighting operators available in the base Studio installation.

YeshSWJTU

Telcontar120 Thank you for your response. I tried SMOTE Upsampling  but it is not supporting to balance multiple minority classes. It works  for binary classification. I will check the weighting operators if it works. thank you

MartinLiebig

Hi @YeshSWJTU,
did you try to set the minority class manually in SMOTE?

Best,
Martin

YeshSWJTU

mschmitz  I tried to set manually but neither it detect the minority class automatically nor it works manually (for binary classification). There is no option for multiple minority classes. Thank you!