Aplicação da análise discriminante linear e redes neurais artificiais para classificação de cerâmicas arqueológicas com base na composição química

The analysis of the microscopic properties of the ceramics has been fundamental for the Archaeological Sciences, since it allows the inference of the cultural framework of the analyzed piece, such as the region of origin of the raw materials, manufacturing technologies and the dynamics of occupation of the spaces. Among the materials found in archaeological excavations, ceramics are the most frequent. Thus, an interdisciplinary program has been established between the Federal Center for Technological Education of Sergipe and the Nuclear and Energy Research Institute with the objective of studying ceramic cultures in the region of Xingo - Sergipe, since there are evidences of ceramics cultures independent of cultures Aratu and Tupiguarani. According to this program, 30 samples of clay and 73 ceramic samples from the Barracao archaeological site were analyzed by Instrumental Neutron Activation Analysis (AANI) to determine the elemental chemical composition of the pieces. It was possible to obtain the concentrations of Na, Lu, Yb, La, Th, Cr, Cs, Sc, Ce, Fe, Eu, Zn, Co, Ta, U and Hf present in the analyzed materials. used to classify the ceramics and clays studied. The discrepant samples were identified by means of the Mahalanobis robusta distance. The interpretation of the database was carried out in two different ways: Linear Discriminant Analysis (ADL) and Artificial Neural Networks (RNA). ADL has been widely used in provenance studies, however, RNAs have been able to overcome some critical points of classical statistical techniques, such as: linearization, influence of noise and discrepant samples. According to the results obtained, the formation of seven compositional groups was observed, being three of clay and four of ceramic. It was also possible to verify that the clay, used in the ceramics analyzed, was not local. Finally, this study makes possible a better understanding of Sergipe prehistory.