Uncovering Regional Structures and Collaboration Hubs in Baden-Württemberg through Clustering a Company Hyperlink Network

Abstract. The hyperlink reference networks of company websites offer a promising approach to modeling inter-firm collaboration and provide valuable insights into e.g. technology diffusion. However, most existing studies analyze these networks without considering the geographic embedding of company locations, disregarding potentially significant spatial factors. To address this gap, we investigate spatial patterns and structures in collaboration networks within the German federal state of Baden-Württemberg. Comparing results from network community detection and unsupervised clustering approaches, we examine firstly, how the collaboration networks are structured in geographic space and secondly, assess whether companies with similar characteristics are also geographically close. Our findings reveal three distinct classes, namely regional, cross-regional, and hub-centered spatial network embeddings, even though a clustering of node attributes did not reveal substantial spatial similarities. These results highlight the important interaction between the virtual hyperlink references of companies and their embedding in geographic space.