一种改进型遗传算法及其在规则提取中的应用

传统的遗传算法在处理复杂的优化问题时容易早熟收敛,陷入局部最优解。为此将免疫原理引入遗传算法,提出了一种新的亲和度定义策略——正弦型亲和度,该策略在对适应度调整时,前期有近似线性的抑制作用,后期则变得平缓。据此设计了一种改进型遗传算法,以提高遗传算法的全局寻优和局部搜索能力。实验结果表明,改进型遗传算法在处理高维多峰函数的收敛速度和收敛精度方面均优于基本遗传算法。以发动机稳定工作状态为例,应用改进型遗传算法实现了飞行状态分类规则的自动获取。测试结果表明,只要训练样本选取得当,获取的规则简洁、有效。

An Improved Genetic Algorithm and Its Application in Rules Extraction

The traditional SGA has the characters of converging early and obtaining easily local best result in the case of processing complicated optimizing problem.So the immune principle is brought in SGA and a new affinity definition strategy(affinity based on sine function) is put forward.This strategy restrains adaptability in the manner of approximate line prophase and flatness anaphase,and a sine based immune genetic algorithm(SIGA) is designed to improve its global and local searching abilities.The experiment results demonstrate that the convergent precision and speed of SIGA is better than those of SGA.Taking the engine stabilization for example,by applying the SIGA,picking up the aviation state classification rules is successfully realized.The testing results indicate that the rule acquired is simple and effective if the training sample is selected properly.