.png){kind=icon}
.png){kind=icon}
Return
An Improved Reference Point-Based Evolutionary Algorithm for Dynamic Multi-Objective Optimization With Preferences
.png){kind=icon}
PRE
AI
DOI：10.1109/TETCI.2025.3611285.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Most existing studies in dynamic multi-objective optimization (DMO) focus on tracking changing Pareto optimal sets and/or the entire Pareto-optimal fronts (PFs). However, there may be scenarios where decision-makers have specific requirements or preferences (e.g., reference points) and are only interested in certain portions of the PF, known as the region of interest (ROI). To address the challenge of preference incorporation in DMO, this paper proposes an improved reference point-based multi-objective evolutionary algorithm with a preference tracking mechanism that simultaneously tracks the changing PFs and ROIs. More specifically, the parameter controlling the size of ROIs in a reference point-based dominance relationship is designed to adaptively adjust with generations for quickly searching the dynamic ROIs. Furthermore, a preference tracking method based on least squares fitting is proposed to predict preferences from historical reference points, enabling the tracking of dynamically changing ROIs. Our proposed method is compared with eight state-of-the-art algorithms on 16 benchmark problems. The empirical results demonstrate the effectiveness of the proposed method both on most test instances and in a real-world application.
Keywords:
Dynamic multi-objective optimization
decision-making
preference prediction
reference points
region of interest
Journal
I
IF:
6.5
Papers:
1.4K
Citations:
4.5K
View more organizations