A STRATEGY OF DIMENSION COMPRESSING IN PATTERN CLASSIFICATIONS AND ITS APPLICATIONS TO PROCESSING COMPLEX CHEMICAL INFORMATION

A new correlative component analysis method was adopted to deal with t he pattern classification problems of the complex chemical information , The classification correlation components obtained from the patterns of high dimensionality are not only mutually independent, but also co ncentrate pattern classification information from the original pattern s. In order to select a reasonable number of components; an algorithm was.proposed in this study, named ratio of correlative index method, i n which the necessary number of components is fixed as the correlative components are calculated. Therefore, the original pattern was transf ormed to a compressed new pattern of lower dimensionality, and the cha racteristics of the new pattern was improved, thus the classification effectiveness was improved either. Finally, multi-classes; example of spearmint essence was applied to check the validity of this new method , The results showed that, the calculation of this new method is reall y convenient and it is an applied method for multi-classes classificat ion problem of complex chemical information backgrounds.