Improved Link Prediction Method for Maritime Silk Road Shipping Network Using Composite Index

Abstract

The container shipping network has evolved with the development of trade, but the mechanism of its evolution is still not well studied. Link prediction model is one of the main methods used to study the network evolution. The classic indicators of the link prediction model only consider the structural characteristics of the network, but there are also real factors that can influence the evolution of the network. Therefore, this paper considers the structural characteristics of the network as an endogenous index, which includes common neighbors, Adamic–Adar, resource allocation, and so forth. Considering the factors that may affect the link between the ports, such as sailing distance, operation, economy, and political factors, this paper first develops different attractiveness models and determines an optimal attractiveness model in case analysis. After that, this paper proposes a composite index that combines endogenous and attractiveness indices. Taking the Maritime Silk Road (MSR) shipping network as an example, an optimal attractiveness index and then an optimal composite index are determined. It was found that the prediction accuracy under the composite index is improved compared with single indices, which confirms the effectiveness and superiority of the proposed index. Finally, from the perspective of the overall network, this paper discusses the possible future routes of the shipping network with the aim of providing a reference for the improvement of connectivity of the shipping network and planning new routes for shipping companies.