dprep (version 2.1)

sffs: Sequential Floating Forward Method

Description

This function selects features using the sequential floating forward method with lda, knn or rpart classifiers.

Usage

sffs(data, method = c("lda", "knn", "rpart"), kvec = 5, 
repet = 10)

Arguments

data

Dataset to be used for feature selection

method

String sequence representing the choice of classifier

kvec

The number of nearest neighbors to be used for the knn classifier

repet

Integer value representing the number of repetitions

Value

• fselect
  a list of the indices of the best features

Details

The Sequential Floating Forward selection method was introduced to deal with the nesting problem. The best subset of features, T, is initialized as the empty set and at each step a new feature is added. After that, the algorithm searches for features that can be removed from T until the correct classification error does not increase. This algorithm is a combination of the sequential forward and the sequential backward methods. The "best subset" of features is constructed based on the frequency with which each attribute is selected in the number of repetitions given. Due to the time complexity of the algorithm its use is not recommended for data sets with a a large number of attributes(say more than 1000).

References

Pudil, P., Ferri, J., Novovicova, J., and Kittler, J. (1994). Floating search methods for feature selection with nonmonotonic criterion function. 12 International Conference on Pattern Recognition, 279-283. Acuna, E , (2003) A comparison of filters and wrappers for feature selection in supervised classification. Proceedings of the Interface 2003 Computing Science and Statistics. Vol 34.

Examples

#---- SFFS feature selection using the knn classifier ----
data(my.iris)
sffs(my.iris,method="knn",kvec=5,repet=5)