Abstract
A speed of large computers has increased a great deal in these days, however, it still takes times to find medians of large networks. This paper aims to examin an improved algorithm for a median analysis, as well as the pruning algorithm for a shortest path analysis it uses. Large directed networks for location models are assumed for the study. When one analyzes a shortest path using the label setting algorithm, without finding the shortest path tree to all nodes in a network, it is possible to know correct node potentials for every node to which permanent lable is set. Taking advantages of this feature, the improved median algorithm is designed to save useless operations following the next : 1) Select a starting node, start a shortest path analysis. 2) Cummulate weighted node potential, every time when a permanent label is set to some node. 3) If a cummulated weighted node potentials exceeds to those calculated for other starting nodes in the prior steps, stop further calculation for this starting node and then move to other starting nodes. Computer experiments proved that the improved median algorithm shortens a cpu time approximately 30-50%, and that the pruing algorithm shortens it approximately 25%.
