View in EDS

EnsembleFL: A Dynamic Multi-Objective Evolutionary Algorithm based on an Ensemble Feed-forward Prediction.

Saved in:
Bibliographic Details
Title: EnsembleFL: A Dynamic Multi-Objective Evolutionary Algorithm based on an Ensemble Feed-forward Prediction.
Authors: Quanheng Zheng1 18839555678@163.com
Source: IAENG International Journal of Computer Science. Aug2025, Vol. 52 Issue 8, p2623-2636. 14p.
Subject Terms: Multi-objective optimization, Evolutionary algorithms, Mathematical optimization, Pragmatism
Abstract: Numerous decision-making challenges across industrial domains can be formulated as Dynamic Multi-Objective Optimization Problems (DMOPs). As an effective computational paradigm for DMOPs, Dynamic Multi-Objective Evolutionary Algorithms (DMOEAs) have attracted significant attention from both academia and industry. To address diverse environmental change patterns, this study proposes EnsembleFL, a novel DMOEA incorporating an ensemble feed-forward prediction mechanism that captures heterogeneous movement patterns of optimal solutions and accurately predicts them in new environments. Experimental evaluations on the CEC'2018 benchmark suite demonstrate EnsembleFL's superior performance compared to five state-of-the-art DMOEAs. Under severe environmental changes, EnsembleFL achieves the best mean Modified Inverted Generational Distance (MIGD) and Modified Hypervolume Difference (MHVD) values on 10 and 13 DMOPs, respectively. In scenarios with mild environmental changes, it attains optimal mean MIGD and MHVD metrics on 5 and 12 DMOPs, respectively. These results validate EnsembleFL's robustness in handling both abrupt and gradual environmental transitions, establishing it as a competitive solution for real-world dynamic optimization challenges. [ABSTRACT FROM AUTHOR]
Database: Supplemental Index
Description
Abstract:Numerous decision-making challenges across industrial domains can be formulated as Dynamic Multi-Objective Optimization Problems (DMOPs). As an effective computational paradigm for DMOPs, Dynamic Multi-Objective Evolutionary Algorithms (DMOEAs) have attracted significant attention from both academia and industry. To address diverse environmental change patterns, this study proposes EnsembleFL, a novel DMOEA incorporating an ensemble feed-forward prediction mechanism that captures heterogeneous movement patterns of optimal solutions and accurately predicts them in new environments. Experimental evaluations on the CEC'2018 benchmark suite demonstrate EnsembleFL's superior performance compared to five state-of-the-art DMOEAs. Under severe environmental changes, EnsembleFL achieves the best mean Modified Inverted Generational Distance (MIGD) and Modified Hypervolume Difference (MHVD) values on 10 and 13 DMOPs, respectively. In scenarios with mild environmental changes, it attains optimal mean MIGD and MHVD metrics on 5 and 12 DMOPs, respectively. These results validate EnsembleFL's robustness in handling both abrupt and gradual environmental transitions, establishing it as a competitive solution for real-world dynamic optimization challenges. [ABSTRACT FROM AUTHOR]
ISSN:1819656X