Attribute selection methods for filtered attribute subspace based bagging with injected randomness (FASBIR)
IM Whittley, AJ Bagnall, Larry Bull, M. Pettipher, Matthew Studley, Firat Tekiner · 2005
Filtered Attribute Subspace based Bagging with Injected Randomness (FASBIR) is a recently proposed algorithm for ensembles of k-nn classifiers [28]. FASBIR works by first performing a global filtering of attributes using information gain, then randomising the bagged ensemble with random subsets of the remaining attributes and random distance metrics. In this paper we propose two refinements of FASBIR and evaluate them on several very large data sets.