An effective ensemble framework for Many-Objective optimization based on AdaBoost and K-means clustering

During multiobjective evolutionary algorithm (MOEA) evolution, the mating and environmental selection op-erators are crucial in selecting promising individuals and enriching the MOEA's performance. However, MOEAs must combat obstacles while addressing many-objective optimization problems (MaOPs). To enhance MOEA performance on MaOPs, various strategies were proposed for mating and environmental selection operators. These strategies were associated with distinct benefits and drawbacks. Therefore, we present an ensemble approach combining mating and environmental selection operators of different MOEAs using an AdaBoost-inspired competitive framework and a K-means clustering-based multistage cooperative framework. In the competitive framework, mating operators compete for resources and preferences assigned to each operator using the AdaBoost strategy. A multistage evolution process is employed, where environmental selection operators collaborate effectively. The K-means clustering algorithm is adopted to select elite individuals for subsequent iterations. K-means clustering requires prior information regarding the number of clusters and is effectively addressed. The proposed ensemble framework's performance is evaluated on 22 benchmark problems with objectives ranging from 5 to 20, comparing it with seven state-of-the-art algorithms. In addition, the MSEMOEA approach is applied to solve three real-world many-objective applications to demonstrate its efficiency. The experimental results demonstrate that the proposed approach achieves better performance than state-of-the-art schemes for convergence and diversity.

Automated summary

An ensemble approach combining mating and environmental selection operators of different MOEAs using AdaBoost and K-means clustering algorithms is proposed to enhance the performance of MOEAs on MaOPs.