Computer Engineering and Applications ›› 2020, Vol. 56 ›› Issue (21): 47-53.DOI: 10.3778/j.issn.1002-8331.1912-0067

Previous Articles     Next Articles

Improved Aggregation-Tree-Based Objective Reduction Optimization for Many-Objective Optimization

WU Tianwei, AN Siguang, SUN Qiqu, LI Mei, SUN Lihong, SHENTU Nanying   

  1. College of Electrical and Mechanical Engineering, China Jiliang University, Hangzhou 310018, China
  • Online:2020-11-01 Published:2020-11-03



  1. 中国计量大学 机电工程学院,杭州 310018


The objective-reduction optimization algorithm solves the problem of consuming too much time by removing or fusing redundant objectives for many-objective optimization problems in practical applications, but the distribution performance of the algorithm will also decline at the same time. Aggregation-tree algorithm defines non-parametric rank conflict and can calculate the conflict value among objectives quickly. However, the robustness of aggregation-tree algorithm needs to be improved, and users need to make decisions to remove redundant objectives by themselves. To solve these problems, an array stacking mechanism is proposed, it defines the trend of conflict and the error of conflict value, and improves the robustness of the algorithm. Objectives combining reduction is proposed by combining redundant objectives with low conflict value, and the cut off conflict value is defined. It is combined with NSGA-III to complete objective reduction optimization for many-objective problems. In order to test the performance of the algorithm in this paper, the optimization of DTLZ test function is compared with other classical many-objective optimization algorithms. The experimental results show that the proposed algorithm has a great advantage in running time, and also ensures excellent distribution performance and convergence performance.

Key words: many-objective optimization, aggregation-tree, conflict value



关键词: 高维多目标优化, 聚合树算法, 冲突度