层级模糊推理系统的递进式建模方法  

A progressive modeling method for hierarchical fuzzy inference systems

作　　者：李玉倩 张大庆[1] 黄胜绢[1] LI Yuqian;ZHANG Daqing;HUANG Shengjuan(School of Science,University of Science and Technology Liaoning,Anshan 114051,China)

机构地区：[1]辽宁科技大学理学院,辽宁鞍山114051

出　　处：《辽宁科技大学学报》2022年第4期300-307,共8页Journal of University of Science and Technology Liaoning

基　　金：国家自然科学基金(61773013)。

摘　　要：在层级模糊推理系统的实际应用中,经常会遇到数据维数较高的问题。本文提出一种递进式层级模糊推理系统建模方法,利用减法聚类构建列联表,通过独立性检验和惯量排序对每一层子系统的输入变量进行选择,当系统精度达到阈值时停止子系统的构建。该方法能够在保持系统精度的同时减少规则数。在层级模糊推理系统建模过程中,保持减法聚类对论域划分,避免建模过程中的重复聚类,降低计算复杂度。利用公测数据集进行仿真计算,并与已有算法进行比较,表明该方法精度良好,且规则数量大幅减少,具有普遍适用性。In the practical application of hierarchical fuzzy inference systems,the problem of high dimension-al data is encountered frequently.In this paper,a progressive modeling method is proposed for hierarchical fuzzy inference systems.The method builds contingency tables using subtraction clustering,and selects input variables of each sub-system through independence test and inertia sorting.The sub-system construction pro-cess stops when the system precision reaches the threshold value.The method can reduce the number of the rules while maintaining the system accuracy.In the subsequent modeling process,the domain division results by the previous subtraction clustering is maintained,which avoids repeated clustering and reduces the compu-tational complexity.The experimental results show that the proposed method performs well and has universal applicability.

关 键 词：层级模糊推理系统 T-S模糊推理系统 减法聚类 惯量 独立性检验 

分 类 号：O159[理学—数学]