Computational considerations in building inter-firm networks

We rarely associate social networks with the movement of freight vehicles. Yet, taking a network perspective on supply chains has seen a strong interest in recent literature. It allows for a variety of system-level analysis that is not possible when taking a single focal firm view as is often the case in more classical supply chain approaches. Creating the network of connectivity on which the analyses are based can be quite a daunting and computationally challenging task. In this paper we create a large-scale network from the movement of commercial vehicles in a metropolitan area in South Africa, using the direct trip between consecutive facilities as a proxy for a tie, or dyad, in the network. We analyse how density-based clustering parameters influence the completeness of the network—that is the number of nodes included—as well as the computational burden of extracting the network. The results of the multi-objective analysis confirm the sensitivity of the resulting network, and suggest much smaller search radii and fewer points per cluster. We also report on a number of node- and network-level properties of the complex network using the proposed clustering configuration on the Nelson Mandela Bay Metropole network in South Africa.