A performance analysis of multi-objective evolutionary algorithms for optimization

Abstract

The scientific objective of the dissertation is to improve the understanding of how evolutionary algorithms work in finding efficient solutions to multiobjective optimization problems through experimental research. The objective of the study is twofold: (1) to describe the performance of evolutionary algorithms in terms of stability, computational complexity, diversity and optimality of solutions in different multiobjective optimization problems, and (2) to describe their strengths and weaknesses in each of the MOOP considered in the study and identify why the MOEA succeeded or failed. The thesis evaluated the performance of two multiobjective evolutionary algorithms on four problem sets that have different search spaces and data structure. The outputs of both MOEAs in each problem set were compared either to other algorithms or with each other, and their results with respect to each problem set were explained. The sensitivity analysis measured the effects of the input parameters on the outputs to describe stability. The multicriteria performance analysis evaluated the quality of nondominated sets in terms of diversity and optimality.