Электронный каталог

Статья

Badawy, W.M.
Neutron Activation Analysis and Machine Learning Models for Elemental Characterization of Archaeological Pottery / W.M.Badawy, M.V.Bulavin, [a.o.]. – Text : electronic // The European Physical Journal Plus. – 2026. – Vol. 141, No. 4. – P. 379. – URL: https://doi.org/10.1140/epjp/s13360-026-07566-z. – Bibliogr.: 67.

This research provides a comprehensive elemental analysis of archaeological pottery by integrating instrumental neutron activation analysis (INAA) and X-ray fluorescence (XRF) with advanced machine learning algorithms, including support vector machine (SVM), random forests (RF), gradient boosting (GB), and multilayer perceptron (MLP). The study aimed to determine the provenance of ceramic sherds and establish geochemical background values using statistical methods, such as interquartile range (IQR), median absolute deviation (MAD), cumulative distribution function (CDF), and Bayesian inference with Markov Chain Monte Carlo (BIM/MCMC). A dataset of 149 pottery fragments from multiple archaeological sites was analyzed, identifying 29 elements. Outliers were filtered to improve data reliability. Fe, K, and Ca were the most frequently detected elements, and the overall elemental profiles closely matched the Upper Continental Crust (UCC) and other geological standards. Machine learning models effectively classified samples and removed anomalies, increasing the accuracy of provenance assignments. Geochemical background values for 25 elements in Bolgar ceramics were established, highlighting Cr, Sb, Mn, As, and Ni as key geochemical markers. These findings provide essential reference data for future geochemical provenance studies and advance the understanding of the origins and distribution of Bolgar pottery.



ОИЯИ = ОИЯИ (JINR)2026