Solving uncapacitated hub location problem using Hopfield-Tank type artificial neural networks

Merkez üslerin yerleşim noktalarının belirlenmesi ve merkez üsler ağının tasarlanmasını içeren merkez üslerin konumlandırılması problemi konum teorisi alanında yaygın olarak çalışılmaktadır. Bu çalışmada, kapasite sınırlamasının olmadığı, sabit konumlandırma maliyetlerinin olduğu, tek tahsisli merkez üsleri konumlandırma problemi için yapay sinir ağı (YSA) temelli bir çözüm yöntemi önerilmiştir. Ayrıca tavlama benzetimine dayanan bir yöntem de geliştirilmiştir. Yöntemin etkinliğini test etmek için, literatürde sıklıkla kullanılan test veri seti kullanılarak, YSA temelli yöntem için elde edilen sonuçlar ile literatürde yer alan en iyi çözümler ve tavlama benzetiminden elde edilen sonuçlar karşılaştırılmıştır.Anahtar Kelimeler: Merkez üsler, yapay sinir ağları, Hopfield-Tank, tavlama benzetimi. In many transportation and telecommunication networks, the cost of carrying a unit of traffic between two points decreases as the capacity of the connection joining the two points increases. It is possible to facilitate this connection by building dedicated channels between each pair of nodes that communicate with each other. However, this would result in higher costs. Because of this fact, it is often convenient to design networks in which traffic is concentrated on high capacity links, even if this traffic travels longer distances. In order to facilitate the flow of the traffic between nodes so as to decrease the overall cost of transportation, some centers known as hubs are introduced. Airline passenger flow, cargo or postal delivery networks, large telecommunication networks are examples of networks utilizing hubs. The problem addressed in this study is the uncapacitated single allocation hub location problem (USAHLP) in which, given n interacting nodes in a network, hubs are fully interconnected and each spoke is assigned to a unique hub. In this study, a solution method based on an artificial neural network framework for the USAHLP is introduced. The heuristic based on simulated annealing is also developed. To present its effectiveness, the solutions of this ANN-based method is compared with the best solutions presented in the literature and the solutions of simulated annealing based heuristic by considering CAB data set. Keywords: Hub location, artificial neural network, Hopfield-Tank, simulated annealing