Classification of Pottery Shards from Diverse Geographical Regions Based on XRF Profiles

Abstract

Pottery fragments found in archaeological sites across the world provide insights into the prevalent manufacturing technology, commercial usage of wares and the socio-politico-economic fabric of the societies which crafted them. Their chemical profile can be used to characterise the clay used in their making, and thus, locate their origins. The technologies for the generation of such geochemical profiles have been around for decades, and several researchers have published the results for their samples. However, such data has undergone just basic statistical analysis. This work collates such data from multiple sources and performs a comparative analysis of multiple machine learning classifiers, to showcase the potential utility of bringing up such datasets for further exploration. It can speed up the segregation and mapping of historical artefacts and add value to archaeological teams working in developing countries of Asia and Africa.