Mutihac Lucia, Mutihac Radu. Mining in chemometrics

Article published in Analytica chimica acta, V. 612, 2008, P. 1-18.Some of the increasingly spread data mining methods in chemometrics like exploratory data
analysis, artificial neural networks, pattern recognition, and digital image processing with
their highs and lows along with some of their representative applications are discussed.
The development of more complex analytical instruments and the need to cope with larger
experimental data sets have demanded for newapproaches in data analysis, which have led
to advanced methods in experimental design and data processing. Hypothesis-drivenmethods
typified by inferential statistics have been gradually complemented or even replaced
by data-driven model-free methods that seek for structure in data without reference to the
experimental protocol or prior hypotheses. The emphasis is put on the ability of data mining
methods to solve multivariate–multiresponse problems on the basis of experimental data
and minimal statistical assumptions only, in contrast to classical methods, which require
predefined priors to be tested against some null-hypothesis.