基于批量模糊学习矢量量化的模糊系统辨识

提出一种基于批量模糊学习矢量量化的模糊系统辨识方法．首先通过优化方法自动调整模糊指数，使所得到的模糊规则前件隶属度函数与聚类规则得到的隶属度函数相比具有较好的可解释性；然后针对模糊系统可解释性与精度之间的困境问题，为保证参数的可理解性．利用带约束的非线性优化方法调整后件参数．并用调整参数的界评估因优化造成参数恶化的程度．仿真实验表明，利用该方法得到的模糊系统模型具有较高的透明度，满足合理的精度．

Identification of fuzzy systems based on batch fuzzy learn vector quantization

An approach to identification of fuzzy system is proposed based on batch fuzzy learn vector quantization to fuzzy modeling. Optimization algorithm is employed to tune the weighting exponent which determines the fuzziness, so that the resulting membership functions are more interpretable than those derived by fuzzy clustering. By considering the dilemma between interpretability and approximation, constrained nonlinear optimization is applied to maintain the interpretation of consequent parameters during tuning. The constraints are implemented by means of limit bounds on updating parameters in order to evaluate the deterioration. Simulation results show that the fuzzy model generated by the proposed method is transparent and reasonably accurate.