论文部分内容阅读
对于经典的TSK模糊系统,模糊规则往往存在冗余而且后件参数缺乏可解释性.本文将TSK模糊系统模型转化为中心化形式,并将其建模过程转换为一个分块稀疏表示问题,提出FCA-sparse CTSK模糊模型.首先运用模糊聚类算法(FCA)对样本特征进行化简,并产生模糊系统字典;再利用存在于中心化TSK(CTSK)模糊模型中的分块结构信息,选取重要的模糊规则并对所选模糊规则的后件参数进行估计.该模型通过中心化方法提高了模糊模型的可解释性,并对模糊规则及模糊规则数同时化简,在合成数据集和真实数据集上都表现出较好的性能.
For classical TSK fuzzy systems, the fuzzy rules are often redundant and the consequent parameters are not interpretable.In this paper, the TSK fuzzy system model is transformed into a centralized form and its modeling process is transformed into a block-sparse representation. FCA-sparse CTSK fuzzy model.Firstly, the fuzzy clustering algorithm (FCA) is used to simplify the sample features and generate fuzzy system dictionaries.Secondly, using the block structure information existing in the centralized TSK fuzzy model, we select the important , And estimate the consequent parameters of the selected fuzzy rules.The model improves the interpretability of the fuzzy model through the centralization method and simplifies the fuzzy rules and the fuzzy rules at the same time. In the synthesis of the data set and the real data Set all showed better performance.