基于聚类分析的模糊分类系统构造方法

提出一种新的利用样本数据构造模糊分类系统的方法.首先对每一类样本进行聚类分析,提出一种自适应确定各类别聚类数目的迭代算法,从而实现对特征空间的划分.然后对每个特征子空间产生一条模糊规则,将所有的规则组合在一起形成初始模糊分类系统,并对该系统进行结构简化和参数优化,在系统结构尽可能简单的前提下,进一步提高系统的分类性能.最后利用该方法对二维特征空间的两类样本和Iris数据样本进行分类.仿真结果表明,该方法能利用较少的模糊分类规则达到较高的识别率.

Approach to Construct Fuzzy Classification System with Clustering

An approach is proposed to construct fuzzy classification systems from the sample data, in which fuzzy cluster method is adopted to model each class. The relationship matrix is defined to evaluate the clustering result and to determine the clusters of each class. A fuzzy rule is created for each cluster and these rules form the rule-based model for classification. Finally, membership functions simplification and parameters optimization are performed to improve the performance of the system. Simulation results show the effectiveness of the proposed method.