Enhancing Efficiency of the Gradual Refinement Approach

The gradual refinement approach has proved to be very efficient tool for finding a good approximation of the Pareto front consisting of various public service designs. The individual designs are represented by sets of p facility locations selected from m possible facility locations. Quality of a design is evaluated according two conflicting criteria, where one of them minimizes system disutility perceived by an average user and the second criterion minimizes disutility perceived by the most exposed users’ minority. The Pareto front consists of such designs that any other design is dominated by at least one of the Pareto front elements. The process of gradual refinement is based on step-by-step updating a current set of non-dominated system designs, when each step represents a neighborhood search procedure performed with one of the current set elements. Repeating of the process proved to be ineffective in comparison to the first application. This contribution is focused on enhancing efficiency of the gradual refinement by a special choice and treatment of the starting design after the first run of the original refinement process is performed.