基于改进萤火虫算法的模糊Petri网学习能力研究

Research on fuzzy petri net learning ability based on improved firefly algorithm

模糊Petri网(Fuzzy Petri net,FPN)被广泛运用于专家系统的建模和推理中。与其它模糊系统建模工具类似,FPN也具有自学习能力不强的固有缺陷。这个缺陷具体表现在三大参数值(权值、阈值和确信度)的确定上,这些参数只能由特定领域的专家结合其经验给出,无法通过机器学习等手段获得。因此如何提高FPN的学习能力来优化其初始的参数值是FPN研究领域的一个研究热点。本文通过对经典萤火虫算法的系统分析,从求解的精度提升和收敛速度优化提出一种改进萤火虫算法,并结合FPN的固有特点,提出一种基于改进萤火虫算法的FPN学习能力训练策略。仿真实验表明,针对同一个FPN模型,经改进后的萤火虫算法训练出来的参数整体性能较其它两类对比算法的训练结果更佳,泛化能力更强。

Fuzzy Petri Net(FPN) is widely used in the modeling and reasoning of expert system. Similar to other fuzzy system modeling tools, FPN also has the inherent defect of weak self-learning ability. This defect is embodied in the determination of three parameter values(weight, threshold and degree of certainty), which can only be given by experts in a specific field combined with their experience and cannot be obtained by means of machine learning. Therefore, how to improve FPN's learning ability to optimize its initial parameter value is a research hotspot in the field of FPN. Based on the systematic analysis of the classical firefly algorithm, this paper proposes an improved firefly algorithm from the perspective of the improvement of solving accuracy and the optimization of convergence speed, and combines with the inherent characteristics of FPN, and proposes a FPN learning ability training strategy based on the improved firefly algorithm. The simulation results show that the overall performance of the parameters trained by the improved firefly algorithm for the same FPN model is better than the training results of the other two algorithms, and the generalization ability is stronger.