An Efficient Approach for Solving Hub Location Problems Using Network Autocorrelation Structures

The properties of spatial information have been shown to aid in identifying optimal solutions for location–allocation problems. Little effort, though, has been made to develop a spatially informed approach to solving hub location problems, as this class of problems entails a more complex model structure and greater challenges in terms of solving capability. To address this issue, this research proposes the spatially informed hub location problem (SI-HLP), derived from investigating the behavior of hub location problems in determining hubs and their allocations to nonhubs to achieve optimal solutions leveraged by underlying spatial characteristics among nodes, links, and routes. The performance of SI-HLP is achieved with two strategies to distinguish essential and nonessential decision variables for location and allocation decision variables, using an innovative convex-hull-based method, HUBI-COV, to capture nodes with high positive network autocorrelations and their allocated links. Simulation experiments under robustly designed settings were conducted to generalize the findings and assess the effectiveness of SI-HLP, indicating that SI-HLPs provide a novel avenue for advancing the solution of large-scale hub location problems.