Feature Selection Based on the Discriminatory Power of Eigenvectors

The necessity of feature selection and feature extraction is well accepted in the field of data processing. Clinical chemistry and haematology stand also in this need but the required multivariate statistical techniques were not introduced properly. A major drawback is that the considered testresults have to be applied in a descriptive as well as discriminatory way. Therefore a single straightforward statistical approach is lacking. So a new concept, the DISCRIMINATORY POWER of EIGENVECTORS, was designed to suit the multivariate statistical techniques of principal component analysis and discriminant analysis at one go to purposes in the medical field. The syndroms of acute and chronic pancreatitis were used to illustrate the proposed method.