A Gene Recombination–Based Path Selection Optimization

Dealing with premature convergence and limited global search efficiency in gene recombination path optimization remains an important challenge. Genetic algorithms can yield satisfactory solutions to such tasks, if it is provided that the selected crossover and mutation operators are well suited, the population is adequately large, and sufficient generations are allowed for the search. Despite various recombination operators, the selection is frequently conducted arbitrarily, which lacks a comprehensive understanding of their capabilities. This paper integrates cyclic shift, shuffle shift, and binary crossover operators, complemented by a dynamic mutation mechanism informed by reinforcement learning and an AMGET explosion module designed to circumvent local optima. Experiments on real BRCA1 gene data demonstrate that the proposed framework converges faster and achieves higher accuracy than conventional genetic algorithms, while remaining robust in large-scale, complex scenarios. Quantitative convergence curves and generation-to-generation fitness variation metrics clearly validate the faster convergence of the proposed method. The approach offers a versatile and effective optimization tool for genomics and bioinformatics research.