The integrated strategy of pattern classification and its application in chemistry

In this paper, the advantages and disadvantages of multivariate discriminant analysis and feed-forward neural networks as classifiers are discussed, and then the strategy integrating the classification methods of two different kinds is proposed. The integrated strategy can be described as follows: use the transform which derives from neural networks to convert the primary matrix which is compose of specimens into a new matrix, and then extract useful components from the converted matrix applying a statistical method, and lastly process the extracted components to establish a discriminant model. Two classifiers, RBF.T-CCA-Fisher and WS.T-CCA-Bayes, which are based on this integrated strategy, are designed to apply respectively to two classification problems: the grade of nature spearmint essence (NSE) and the classification of toxicity of amines. The results show that the new classifiers have the better classification correctness and wider applicability than any statistical method or neural network does.