Improving an immune-inspired algorithm by linear regression: A case study on network reliability

• Published in: Knowledge-based systems Vol. 299; p. 112034
• Main Authors: Cutello, Vincenzo, Pavone, Mario, Zito, Francesco
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 05.09.2024
• Subjects: Dynamic mutation rate; Immune algorithm; Machine learning; Metaheuristic; Network optimization; Network reliability maximization; Ordinary least squares regression; Immune algorithm; Network reliability maximization; Metaheuristic; Ordinary least squares regression; Dynamic mutation rate; Network optimization; Machine learning
• ISSN: 0950-7051; 1872-7409
• DOI: 10.1016/j.knosys.2024.112034

Bio-inspired algorithms are widely and successfully used in solving complex optimization problems. However, a common challenge faced by these algorithms is getting stuck in local search spaces and failing to explore new regions to further enhance the solution. Such a limitation is often attributed to the influence of the working parameters that need to be carefully tuned so to obtain an optimal performance. Unfortunately, finding the right parameter settings is problem-specific and can vary from one instance to another. To address this issue, this research investigates the integration of machine learning techniques with bio-inspired algorithms to dynamically adjust algorithm parameters based on solution quality. Specifically, the study focuses on an immune system-inspired algorithm and introduces a new strategy to mutate individuals in order to explore the search space more efficiently. By leveraging prediction techniques such as Linear Regression, the algorithm can predict local optima and adjust its search direction accordingly. The proposed method, referred to as Dynamic-IA, is evaluated using the Network Reliability Problem as a case study. Several network topologies are considered in our experiments and the results obtained by our algorithm when compared with the standard Immune Algorithm and other metaheuristic algorithms prove our strategy. •Immunological Algorithm with dynamic mutation rate.•Combining Immune Algorithm with Machine Learning techniques.•Ordinary Least Squares Regression method to dynamically determine the number of mutations.•The Maximum Relevance Minimum Redundancy algorithm was used to detect the most important parameter of Immune Algorithm.•The Network Reliability Problem was tackled.