A self-evolving fuzzy system online prediction-based dynamic multi-objective evolutionary algorithm

The changes of dynamic multi-objective optimization problems in decision space are usually nonlinear. However, the previous dynamic multi-objective evolutionary algorithms usually use linear prediction models to generate the initial population in the new environment, and some nonlinear prediction mo...

Full description

Saved in:
Bibliographic Details
Published in:Information sciences Vol. 612; pp. 638 - 654
Main Authors: Sun, Jing, Gan, Xingjia, Gong, Dunwei, Tang, Xiaoke, Dai, Hongwei, Zhong, Zhaoman
Format: Journal Article
Language:English
Published: Elsevier Inc 01.10.2022
Subjects:
Dynamic multi-objective optimization
Self-evolving fuzzy system
Time series
Weight vectors
Weight vectors
Self-evolving fuzzy system
Dynamic multi-objective optimization
Time series
ISSN:0020-0255, 1872-6291
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The changes of dynamic multi-objective optimization problems in decision space are usually nonlinear. However, the previous dynamic multi-objective evolutionary algorithms usually use linear prediction models to generate the initial population in the new environment, and some nonlinear prediction models often have high computational cost. Therefore, it is difficult to quickly and accurately respond to nonlinear environmental changes. This paper presents a dynamic multi-objective evolutionary algorithm based on online prediction of self-evolving fuzzy system (SEFS). In this algorithm, the decomposition based multi-objective evolutionary algorithm (MOEA/D) acts as the static optimizer. When the environment changes, individuals are first put into an associate set of their corresponding weight vectors. Then, the time series of each variable is constructed based on the associate set, and the SEFS online prediction model is established. Finally, an environmental response strategy based on SEFS is designed to quickly generate an initial population with high performance in the new environment. The proposed algorithm is compared with seven state-of-the-art dynamic multi-objective evolutionary algorithms on 20 benchmark functions. Experimental results show that the proposed algorithm can quickly and accurately respond to nonlinear environmental changes, and has competitiveness.
ISSN:0020-0255
1872-6291
DOI:10.1016/j.ins.2022.08.072