An improved NSGA2 to solve a bi-objective optimization problem of multi-state electronic transaction network

Abstract

An electronic transaction network (ETN) plays a very important role in communications among trading partners. Transmission reliability is of concern to system supervisors. This study adopts a binary-state physical line allocation strategy, minimizing cost and maximizing transmission reliability for an ETN with a known network structure, in which the ETN is represented by arcs and nodes. The strategy is to allocate adequate binary-state physical lines to arcs. Particularly, the physical lines allocated to the same arc could be in correlated failure owing to maintenance. That is, the ETN can be modeled as a multi-state flow network with correlated failures for reliability evaluation. For solving this bi-objective optimization problem, an improved fast non-dominated sorting genetic algorithm II (iNSGA2), integrating the NSGA2 and k-means algorithm, is proposed, where the k-means is utilized to expand the search space of the NSGA2. A set of non-dominated solutions is found by the iNSGA2, and then, the technique for order preference by similarity to an ideal solution (TOPSIS) is adopted to determine the compromise alternative from the set. By solving this problem, the system supervisor can improve ETN stability at a reasonable expense without changing the network structure.