Community Detection in Archaeology: Testing Robust Algorithms on Early Copper Networks in the Balkan

In collaboration with the Institute of Classical Studies, we are delighted to launch a new seminar series titled Shared Horizons: A Dialogue Between Classical Archaeology and Archaeological Science. This series aims to spark an open and stimulating conversation between archaeologists and archaeological scientists, exploring how diverse approaches — from field excavation to laboratory analysis — can together illuminate the ancient world.

The goal of Shared Horizons is to foster closer dialogue and collaboration between researchers working across different yet complementary areas of archaeology, united by shared questions about material culture, technology, and human experience in the past.

In this next hybrid seminar on Wednesday 10 December from 5pm, Miljana Radivojevic (UCL Institute of Archaeology) will give a presentation entitled Community Detection in Archaeology: Testing Robust Algorithms on Early Copper Networks in the Balkans.

Abstract

The emergence of metallurgy in prehistoric Europe is marked by complex trajectories of copper production and circulation, particularly visible in the Balkans between the 6th and 3rd millennia BC. While provenance studies have long demonstrated the exploitation of major ore sources such as Majdanpek and Ai Bunar, the organisation of these early metal networks has remained insufficiently explored. 

In this study, we apply community detection methods from network science to a compositional dataset of 410 copper-based artefacts from c. 80 sites, spanning c. 3000 years (6200–3200 BC). Using trace element and lead isotope analyses, we construct networks of copper supply and evaluate their modular structure. To ensure robustness, we systematically test seven community detection algorithms, including Louvain, Leiden, Spinglass, and Eigenvector approaches, critically assessing their performance on archaeological compositional data. The resulting communities show strong correlations with traditionally defined archaeological cultures such as Vinča, KGK VI and Bodrogkeresztúr, highlighting the potential of algorithmic methods to reassess the very concept of “archaeological culture.” 

Our results demonstrate that community detection provides a rigorous quantitative framework for studying prehistoric exchange, while also exposing challenges of resolution, reproducibility, and data quality. This case study illustrates both the promise and limitations of applying advanced network science to archaeological big data.

The Shared Horizons series organisers are Maja Miše and Stefania Alfarano (UCL).