Classification Mapping and Its Application on Chemical Systems

In this paper, the conception and the construction of classification maps (CMs) are discussed. Methods called Fp mapping and Sp mapping are proposed based on discriminant analysis (DA) and projection pursuit (PP). By these methods, an experimental data set with classification information can be mapped from multidimensional data space onto a series two-dimensional projection keeping the original classification structure in the process. These projections show intuitively the semiquantitative relationships between the study object and its controlling factors that we cannot observe directly from multivariate data space. Based on existing experimental data, these methods are useful in giving valuable suggestions on further experimental works. As an example, they were applied to the study of a kind of propellant material and a kind of welding material and were proven effective in the optimization of the material's performance.