Discriminant plots obtained via removal of classification structures

A new method for stepwise discriminant analysis, called “removal of classification structures”, is proposed. Its novelty lies in the transformation of the data along a found discriminant direction into data without discriminant information and iteration to obtain the next discriminant direction. The proposed method can be applied to any discriminant criterion. It is free to search for discriminant directions oblique to each other and ensures that the informative directions already found will not be found again later. A comparative experiment with artificially constructed data shows that the new method is better than the stepwise methods used in statistical pattern recognition