Evaluation by pattern recognition methods of laser-micro-spectral analysis data of Roman coins

Data produced by the non-destructive method of laser-microspectral analysis were treated by a pattern recognition method in order to characterize silver Antoniniani coins produced in various mints of the Roman Empire in the middle of the 3rd century. The aim of the work was the classification of the sample coins according to their origin from different mints, and the assignment of some coins of „uncertain origin“ to these classes on the basis of their chemical composition, i.e. the content of the significant minor and trace elements: Au, Zn, Ni, Bi, Sn, Pb. The Karhunen-Loeve transformation and the K-nearest-neighbours pattern recognition methods were applied for processing the analytical data. As a result of the investigations nearly all the coins studied could be assigned to the presumed groups of the mints considered and an objective support was given to numismatists in aid of their coin-classification work.