Shape-based visual analysis of solutions for multiobjective optimization problems
Christian von Lücken, Abrahan D. Fretes Prieto, Arsenio M. Ferreira Vera y Aragon, Benjamı́n Barán · 2017 XLIII Latin American Computer Conference (CLEI) · 2017
In general, Multiobjective Optimization Problems (MOPs) with conflicting objectives requires a Decision Maker to select a solution from a set of alternatives in the Pareto Front. A visual approach is a valuable alternative to analyze the several options that may exist. Thus, to support the visual analysis, this work proposes to combine a clustering method based on the shape of the solutions together with the parallel coordinate graph. The shape of a solution is defined as the positions of the ordered values of the normalized objectives that correspond to a solution. The method proposed in this work, allows exploring groups of solutions with similar shapes, to generate visual summaries with a high degree of representativeness, and to calculate statistical data of the groups and to identify the solutions. Also, this paper presents a software tool called Tava. This tool was developed to perform the visual analysis interactively.