英文摘要：
Fuzzy Petri Net (FPN) can reasonably describe the uncertainty and fuzziness in the real world and can be used for the reasoning of uncertain knowledge, which 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.
|