r/ combinatorial optimization function/package requiring minimal up-front work

Question

I have a data frame consisting of names of preprocessing methods.

*Impute*    *Scale*      
naomit      noscale      
knnimpute   noscale              
naomit      scale        
knnimpute   scale     

In step one function g() executes the methods row-wise to create a preprocessed data set. For first row: identity(na.omit(data))

In step two classification error in computed for each preprocessed data set. The objective is to find a combination that minimizes classification error.

There are thousands of combinations. Currently, I use full blind or simple grid search. I need a more intelligent method to find preprocessed data sets worth testing.

I know there is the CRAN task view for optimization and I have tried to learn conceptual issues from here (http://dl.acm.org/citation.cfm?id=937505).

What would be a good R combinatorial optimization package/ function to find approximately best solution faster with mimimal up-front work?

Answer 1

Answering my own question: made package 'metaheur' for the purpose above.