APPLICATION OF ANALYTICAL TECHNIQUES AND MACHINE LEARNING MODELS FOR THE IDENTIFICATION AND CULTURAL ATTRIBUTION OF ANCIENT POTTERY: A CASE STUDY FROM BOLU CLAUDIOPOLIS, TURKEY
Journal of Ancient History and Archaeology, cilt.13, sa.1, ss.179-192, 2026 (ESCI, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 13 Sayı: 1
- Basım Tarihi: 2026
- Doi Numarası: 10.14795/jaha.13.1.2026.1504
- Dergi Adı: Journal of Ancient History and Archaeology
- Derginin Tarandığı İndeksler: Emerging Sources Citation Index (ESCI), Scopus, Central & Eastern European Academic Source (CEEAS), Directory of Open Access Journals
- Sayfa Sayıları: ss.179-192
- Anahtar Kelimeler: Ancient Pottery, Bolu Claudiopolis, Fuzzy Logic, Machine Learning, XRF Analysis
- Ondokuz Mayıs Üniversitesi Adresli: Evet
Özet
Understanding the origin, manufacturing technology, and cultural affiliation of ancient pottery is a major objective in interdisciplinary archaeological research. In this study, 20 pottery samples from the ancient site of Bolu Claudiopolis (modern-day Bolu, Turkey) were investigated using a multi-analytical approach. X-ray Fluorescence (XRF), X-ray Diffraction (XRD), and Fourier Transform Infrared Spectroscopy (FTIR) were employed to characterize the chemical, mineralogical, and structural properties of the samples. The resulting datasets were analyzed using several machine learning algorithms, including K-Nearest Neighbors (KNN), Support Vector Machine (SVM), and Random Forest (RF), implemented in Weka software. Principal Component Analysis (PCA) was applied for dimensionality reduction and pattern recognition, while fuzzy logic was used as a complementary framework for approximate cultural attribution. In addition, data fusion strategies were applied to combine information obtained from the different analytical techniques. The results revealed compositional variability among the pottery samples and indicated that most of them were more closely associated with Roman compositional patterns, whereas a limited number showed affinity with Greek-related characteristics. Among the tested models, Random Forest provided the most consistent classification performance. XRD and FTIR results further suggested medium firing temperatures and the presence of silicate-based raw materials with specific mineral phases. Overall, this study demonstrates the value of integrating analytical techniques with machine learning and fuzzy logic for the classification and comparative assessment of ancient pottery.