Software defect prediction based on support vector machine optimized by reverse differential chimp optimization algorithm

Lifang Chen

doi:10.1007/s41060-025-00726-x

2025-02-04

PRE

DOI：10.1007/s41060-025-00726-x

原文链接

原文求助

摘要

En 中文

As software development becomes increasingly complex, defect prediction plays a crucial role in ensuring software quality. Existing software defect prediction methods face limitations in parameter selection for classification models, particularly in selecting the penalty factor and kernel function parameters in SVM. Traditional optimization methods often struggle with insufficient convergence precision and a tendency to fall into local optima. To address these issues, this paper proposes the reverse differential chimp optimization algorithm (RDChOA). RDChOA improves upon the traditional chimp optimization algorithm by incorporating the Hammersley sequence initialization, lens-imaging reverse learning strategy, and differential evolution strategy, thereby enhancing global search capability and reducing the likelihood of converging to local optima. RDChOA starts by using the Hammersley sequence to initialize the chimp population, increasing initial population diversity. In the later stages of the algorithm, the lens-imaging reverse learning strategy is employed to update the positions of attackers, further expanding the search space and avoiding local optima. Finally, the differential evolution strategy is applied to adjust the positions of regular chimp individuals, boosting global optimization performance. Through these innovations, RDChOA effectively optimizes the parameters of SVM, addressing the challenges of parameter selection and generalization capability faced by traditional defect prediction algorithms. Experimental results demonstrate that RDChOA performs excellently on eight benchmark test functions, outperforming other swarm intelligence optimization algorithms. Moreover, when applied to multiple public software defect prediction datasets, RDChOA-SVM also shows significant advantages in prediction accuracy.

Keyword:

Reverse differential chimp optimization algorithm

Software defect prediction

Support vector machine

Hammersley sequence

Lens-imaging reverse learning

Differential evolution