NEW CHEMOMETRIC TOOLS TO STUDY THE ORIGIN OF AMPHORAE PRODUCED IN THEROMAN-EMPIRE

Information about geographical and chronological origin is often requi red of archaeological samples. In order to obtain such information, pa ttern recognition techniques are now used as valid tools for processin g series of data from morphological and chemical analyses. This articl e reviews the advantages and disadvantages of artificial neural networ ks (ANNs) methods and compares them with chemometric techniques such a s the standard clustering method, principal component analysis, (PCA) and SIMCA, Kohonen, Back-propagation of errors, and counter-propagatio n learning strategies of ANNs are used to study 160 amphorae dating to the 5th century A.D. The majority (128) of the amphorae are of known and 32 of unknown geographical origin, Some predictions about the unkn own samples are made, thus showing the potential of ANN techniques and their possible contribution to archaeological studies.