Enhancement of Running Time - Forward Elimination
Muhammed_Fatih_
New Altair Community Member
Dear Community,
I have a question with regard to the ForwadElimination subprocess. Which classification model would you use within crossvalidation in order to enhance the running time? I am currently conducting a Forwald Elimination based on matrix with 72.000 rows and 9000 attributes. I've chosen SVM as classifier for my bipartite labeling. The model is still running since 2 days.
Thank you for your answer!
Fatih
I have a question with regard to the ForwadElimination subprocess. Which classification model would you use within crossvalidation in order to enhance the running time? I am currently conducting a Forwald Elimination based on matrix with 72.000 rows and 9000 attributes. I've chosen SVM as classifier for my bipartite labeling. The model is still running since 2 days.
Thank you for your answer!
Fatih
0
Best Answer
-
Hi,
SVM can be quite slow, depending on the number of examples and attributes, and also the parameters (high values of C).
Did you manage to build one model?
10 models per cross validation * 9,000 runs just in the first round * time for building a model on one attribute = runtime of the first round.
You might want to sample your data before doing the feature selection, or use a different method like evolutionary feature selection. You should also check non-model based methods like Remove Correlated Attributes, these can be faster.
Regards,
Balázs1
Answers
-
Hi,
see my answer here: https://community.rapidminer.com/discussion/55878/how-does-feature-selection-forward-elimination-work-in-detail
Regards,
Balázs2 -
Hi,
thank you for your answer. This was again very helpful. Beside that, is there a possibility to estimate the running time for the execution of Foward Elimination using SVM?
Best regards,
Fatih0 -
Hi,
SVM can be quite slow, depending on the number of examples and attributes, and also the parameters (high values of C).
Did you manage to build one model?
10 models per cross validation * 9,000 runs just in the first round * time for building a model on one attribute = runtime of the first round.
You might want to sample your data before doing the feature selection, or use a different method like evolutionary feature selection. You should also check non-model based methods like Remove Correlated Attributes, these can be faster.
Regards,
Balázs1 -
Hi,
what do you mean with evolutionary feature selection - the Optimize Selection (Evolutionary) operator ?
Best regards,
Fatih0 -
2