This paper describes a novel approach of the classification of modified sta
rches with regard to their origin and the recognition of their properties a
nd modifications by chemometric treatments of infrared spectra, FTIR spectr
a of 94 starch samples are recorded with an horizontal ATR accessory Statis
tical (PCA, PLS) and neuronal (Kohonen and feedforward networks) methods ha
ve been used. Unsupervised projections and supervised prediction results ar
e compared.