M. Járó - L. Költő szerk.: Archaeometrical research in Hungary (Budapest, 1988)
Analysis - GEGUS Ernő, BORSZÉKI János: Investigation of archaeological metal findings by a laser-microspectral analysis method and characterization of results using pattern recognition methods
Bi 306.7 and 289.8 Cr 267.7 Co 345.3 Cu 327.4 (copper, bronze, brass: 249.2; 282.4; 296.1; 306.3) Fe 259.9 and 275.5 Mg 279.5 (characteristic for soil) Mn 257.6 Ni 310.2 and 341.4 P 255.3 Pb 283.3 (above 0.5%: 257.7) Pd 340.4 (in gold samples) Pt 306.4 (in gold samples) Sb 259.8 (high Fe content interferes) Si 288.1 (characteristic for soil) Sn 284.0 (above 3%: 278.0) TÍ327.2 V311.0 Zn 334.5 (above 0.5%: 307.6) 3. Pattern recognition in archaeometry The use of methods of natural sciences in archaeological and historical research has led, in the last few years, to the development of the science of archaeometry. The new branch of science applies today not only the instruments of modem instrumental analysis but to an increasing degree, the means of mathematics, too. Among these, the mainly and most effectively used mathematical statistical methods are the pattern recognition procedures. Using these, properties or features in the measured data matrix of the investigated samples that are not directly measurable can be estimated or determined. This possibility means, in archaeometry, that say the origin of the archaeological sample, the place where the coin was minted, thus the age of the sample, can be estimated with good approximation by using pattern recognition data processing of the measurement results. Laser-microspectral analytical results obtained from coins and other metallic findings were first of all processed archaeometrically . In possession of a mathematical statistical data processing system, the higher the number of investigated samples and the more parameters measured (among them an optimal number of characteristics), the more certain is the achieved result. Questions which can be answered by using pattern recognition methods may be divided into two main groups: 1. if no previous information exists about the sample mass, the samples could be divided into groups according to a more definite point of view than if a method of classification were to be applied; 2. another way is that a ranging of unknown sample points will be carried out on the. basis of the measured characteristics, using known classes or groups; in this case one speaks about ranging into classes. In archaeometric investigations, information determined on the basis of a few or primarily of external characteristics is generally available. Then, with the task of deciding whether to classify or to range into classes, the first step of data processing is always the feature selection. During this data preprocessing stage, from the measured characteristics the most important with regard to the expected or probable grouping features are
