Feature-selection overfitting with small-sample classifier design
Academic Article
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
The role of feature selection in overfitting the data and how this is exacerbated by high dimensionality and small samples is described. The best feature must be found from among all possible feature subsets. Multivariate prediction, and error estimation can severely impact feature selection while redundancy. A feature-selection algorithm is part of the classification rule when used. If a feature-selection algorithm reduces the number of variables to m