FEATURE SELECTION USING THE INTEGRAL DOE AND MANOVA TECHNIQUES FOR CLASSIFICATION

摘要

In this paper, we present a new approach for feature selection using an integratedrnDOE and MANOVA technique. The single feature variable selection algorithm isrnadopted to eliminate poor discriminated feature variables. The Plackett-Burman (PB)rnresolution III design is then constructed to select the remaining feature variables. ThernMANOVA technique is used to calculate the Pillai statistic as the response for the PBrndesign of experiment. Statistical analysis is then executed to obtain the optimalrnmultiple feature variables for multiple groups. The Iris data is used as an example tornshow how the proposed analysis procedure can acquire an optimum subset of featuresrnfor classification.