Chemometrics is a method to extract the significant information from a big amaount of data, e. g. hundrets of spectral data points. In most cases these data first have to be separated from disturbances or external distorsions. This is achieved using special chemometric software.
The target is to find a minimum number of factors (loadings) that on one hand contain all the available information, but on the other hand are independent from each other (orthogonal). These then can be used in a regression to develop a model.