Abstract
Accepted: 03.12.2014 Evolutionary algorithms are very popular methods to solve multiobjective optimization problems. In literature, there are many multi-objective evolutionary algorithm methods. It is possible to improve these methods. This paper proposes a line of refinements and modifications on the fitness assignment part of well known SPEA method. The density information generated by k-NN (k-th nearest neighbor method) and the concept of domination power are used to modify SPEA in order to deliver a much refinement fitness values in terms of uniqueness and homogeneity. Two modified methods are proposed (named SPEAmod1 and SPEAmod2). The main goal is to extract more information from a population (and is to deliver to the decision maker) by using proposed variants than the original method. The original method and its modifications are tested on four well-known test functions in the literature. The proposed modifications are shown to be superior to the original method. Finally, fitness assignment capability of original SPEA method is significantly improved.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call