GNGTS 2024 - Atti del 42° Convegno Nazionale

Session 3.2 GNGTS 2024 Results From the cluster analysis applied to the EM data for two operatng frequencies (5 kHz, 10 kHz), it turns out that the Silhouete coefcient always identfes an optmal number of clusters, smaller than that from the Elbow method. In other words, optmal clusters validated by the Silhouete coefcient correspond to areas in the archaeological site that are much larger than those identfed by the Elbow method. Looking at the clusters of major interest for archaeological studies, which are characterized by the highest values of magnetc susceptbility and electrical resistvity, it is found that groups from the Silhouete validaton analysis can contain up to three of the optmal clusters identfed by the Elbow method. However, in some cases the clusters from both the Elbow method and Silhouete coefcient agree in the identfcaton of potental target areas, i.e., where the presence of a ditch is inferred. We stress that all the anomalous areas identfed by the analysis of magnetc data fall within the areas of potental interest for the archaeological excavaton operatons identfed by the present clustering analysis. Hence, the proposed approach shows that the applicaton of unsupervised machine learning techniques to FDEM data is promising in improving the efciency of this method making it more atractve for archaeological research. References Bhatacharya, S.; 2021: A Primer on Machine Learning in Subsurface Geosciences, 1, 1–172. Cella, F., Fedi, M.; 2015: High-resoluton geophysical 3D imaging for archaeology by magnetc and EM data: the case of the iron age setlement of Torre Galli, Southern Italy. Surveys in Geophysics, 36(6), 831-850. Cella, F., Paolet, V., Florio, G., Fedi, M.; 2015: Characterizing elements of urban planning in Magna Graecia using geophysical techniques: the case of Tirena (Southern Italy). Archaeological Prospecton, 22(3), 207-219. Di Maio, R., La Manna, M., Piegari, E.; 2016: 3D reconstructon of buried structures from magnetc, electromagnetc and ERT data: Example from the archaeological site of Phaistos (Crete, Greece). Archaeological Prospecton, 23, 3-13. El-Qady, G., Metwaly, M., Drahor, M. G.; 2019: Geophysical techniques applied in archaeology. Archaeogeophysics: State of the art and case studies, Springer, 1-25. Everet, M.E.; 2013: Near-Surface Applied Geophysics. Pp.209. Corresponding author: angelica.capozzoli@unina.it