FTIR spectra combined with machine learning to reveal the amber trade during the Han dynasty (202 BCE–220 CE)

Abstract

Amber has long been recognized as a key to providing vital insights into the commercial exchanges of ancient civilizations. This study analyzed fifty amber beads from Southern China (Guangdong Province), fifteen from Northwestern China (Qinghai Province), and thirteen from Central China (Hunan Province) during the Han dynasty to trace their origins by Attenuated Total Reflection Fourier Transform Infrared (ATR-FTIR) and Diffuse Reflectance Infrared Spectroscopy (DRIFTS). An automated amber origin classification method was proposed to verify the manually identified amber origin and streamline the analysis. The factors influenced by post-depositional alterations, including impurities, degrees of aging, and testing methods, show no correlation with amber origin, which may diminish classification effectiveness. Each of these factors was examined individually, leading to the identification of credible origin fingerprint peaks in the IR spectra. Valid data processing steps focused on fingerprint peaks’ regions were designed based on the IR spectra features. Density-based spatial clustering of application with noise (DBSCAN) was adopted to discover groups within the data. Classification models were then created using Support Vector Machine (SVM), Decision Tree (DT), and K-Nearest Neighbor (KNN) algorithms. The machine learning (ML) models demonstrated high accuracy comparable to manual identification while also achieving greater efficiency. According to the manual and ML results, the prevalence of both Baltic and Burmese amber in Guangdong and Hunan and the mainstream of Baltic amber in Qinghai were discussed, along with the potential amber trade routes. Our results provided evidence for the vibrant exchange between China, Southeast Asia, Southern Asia, and Central Asia in the Han dynasty and indirect trade links from the Han Empire to ancient Europe.