Memetic algorithm for the balanced resource location problem with preferences

In this paper, a variant of the Load Balanced Resource Location Problem is considered, which arises from network optimization in cases when equity criteria of service providers is important. The problem has numerous applications, for example, in designing telecommunication systems, optimizing Web traffic, server load balancing, Big Data storage and management, etc. In the load balance model considered in this study, we involve users' preferences to be served by a certain resource, and ensure that each user is assigned to its most preferred resource. The goal is to establish a fixed number of resources from the set of potential resource nodes, and to assign each user to its most preferred established resource, such that the difference in the maximal and minimal assignment costs among established resources is minimized. Due to the complexity of the problem, optimal solutions are obtained only for smaller-size problem instances. A Memetic Algorithm (MA) based on hybridization of Evolutionary Algorithm and Local Search method is proposed for solving the problem, especially in the case of a network that involves large number of nodes. Computational results obtained on two data sets show that the proposed MA quickly reaches all optimal solutions obtained by CPLEX solver on smaller-size problem instances, and produces solutions on large problem instances that could not be solved to optimality by CPLEX.