A graph neural networks and improved whale optimization algorithm-based approach for cloud service composition

• Published in: Cluster computing Vol. 28; no. 7; p. 480
• Main Authors: Yin, Zuheng, Xu, Hongzhen, Liu, Aihua
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.09.2025; Springer Nature B.V
• Subjects: Algorithms; Cloud computing; Cognition & reasoning; Composition; Computer Communication Networks; Computer Science; Deep learning; Energy consumption; Fuzzy logic; Graph neural networks; Heuristic; Heuristic methods; Local optimization; Machine learning; Methods; Neural networks; Operating Systems; Optimization algorithms; Processor Architectures; Quality of service; Real time; Semantic web; User needs; Historical service solutions; Graph neural networks; Whale optimization algorithm; Cloud service composition; Quality of service
• ISSN: 1386-7857; 1573-7543
• DOI: 10.1007/s10586-025-05139-w

Cloud service compositions have attracted the attention of many scholars as an important part of the cloud computing field. Currently, efficiently performing cloud service composition while meeting quality of service requirements has been one of the research challenges in service computing. On the one hand, cloud service composition algorithms that find optimal solutions are inherently computationally intensive. As a result, it is often impossible to generate optimal solutions in real-time when faced with an ever-increasing number of cloud services. On the other hand, while meta-heuristic algorithms can handle large and complex cloud service spaces and generate solutions quickly, the algorithms are prone to local optimization and have difficulty finding globally optimal solutions. The question of balancing computational efficiency and optimality in the composition of cloud services has always been one that needs to be addressed. In this paper, we propose a cloud service composition method based on graph neural networks and an improved whale optimization algorithm. First, we use a graph neural network to mine the potential relevance of historical service solutions and predict the probability of each cloud service constructing the corresponding solution for a specific task. Secondly, an improved whale optimization algorithm based on diversity variance operation and nonlinear convergence factor subject to chaotic perturbation is proposed to find the optimal service composition scheme for high-probability cloud services. Finally, experiments show that our method improves solution quality by at least 3.3% and up to 11.3% compared to other excellent algorithms.