Minimax Zoning Technique for Contiguous Clustering: Application to Determine Delivery Zones

A distribution company plans on improving its overall delivery scheme. However, the first phase in this series of changes involves the task of zoning the company’s Metro Manila clients taking into consideration three factors: (1) contiguity; (2) composition of orders; and (3) frequency of orders. The first consideration, contiguity, is difficult to account for because of lack of available techniques in literature that could adequately model it. Thus, a new approach, the Minimax Zoning Technique, was explored. This technique made use of the minimal spanning tree formulation – which was first solved conventionally with software, then, applying a set of criteria, was stripped of heavy branches. In this approach, contiguity was modeled by making connections possible only between adjacent client accounts, which acted as the nodes in the model. Meanwhile, the weights of the branches were based on the similarity of two connected squares as measured by the differences in the second and third factors. Performing the algorithm on the case resulted to a configuration composed of five exclusive zones. Due to some weaknesses in the algorithm, optimality could not be guaranteed. Nevertheless, the results achieved were deemed satisfactory based on statistical comparisons made with other zoning configurations.