Abstract

Interactive multiobjective optimization (IMO) methods aim at supporting human decision makers (DMs) to find their most preferred solutions in solving multiobjective optimization problems. Due to the subjectivity of human DMs, human fatigue, or other limiting factors, it is hard to design experiments involving human DMs to evaluate and compare IMO methods. In this paper, we propose a framework of a virtual-DM library consisting of a variety of virtual DMs which reflect characteristics of different types of human DMs. The virtual-DM library is used to replace human DMs to interact with IMO methods. The virtual DMs in the library can express different types of preference information and their most preferred solutions are known. When interacting with an IMO method, the library can select an appropriate virtual DM to provide preference information that the method asks for based on solutions offered by the method. Four types of hybrid virtual DMs are constructed to emulate human DMs with different personalities and dynamically changing preference structures. They can be used to test the ability of IMO methods to adapt to different human DMs and capture DMs' preferences. The usage of these four types of virtual DMs are demonstrated by comparing two IMO algorithms.

Full Text
Published version (Free)

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