/qsm-advanced-feature-selection

QSM feature selection from high-dimensional feature spaces

Primary LanguageMATLABGNU General Public License v3.0GPL-3.0

DIE1VA

DOI

Decimated Input Ensembles for One Vs All classification (DIE1VA) is a filter-wrapper hybrid method for selecting features to be used in One-vs-All classification. The method implements majority voting to combine the predictions of a user-defined number of One-vs-All classifiers for each class that use different feature sets. Example.mlx contains a Matlab live script example of using DIE1VA after some preprocessing. Feature values (FV) and names (FN) used in the example script can be acquired by executing compute_features.m on some QSMs. Refer to the paper "Feature selection in high-dimensional feature spaces for tree species classification from quantitative structure models" or the help documentation of a particular function for further details.