摘要
提出一种基于模糊聚类和遗传算法的模糊分类系统的设计方法.首先定义了模糊分类系统的精确性指标,给出解释性的必要条件.然后利用聚类有效性分析确定模糊规则数目,利用模糊聚类算法辨识初始的模糊分类系统.随后利用模糊集合相似性分析与融合对初始的模糊分类系统进行约简,提高其解释性;利用遗传算法对约简后的模糊分类系统进行优化,提高其精确性,该过程反复迭代直至满足中止条件.最后利用该方法进行Iris数据样本分类,仿真结果验证了该方法的有效性.
An approach of constructing interpretable and precise fuzzy classification system based on fuzzy clustering and genetic algorithm is proposed. First, the precision index is defined, and the necessary conditions of interpretability are analyzed. Second, the number of fuzzy rules is determined by cluster validity indices, and the initial fuzzy classification system is identified using a fuzzy clustering algorithm. Subsequently, the method of merging similar fuzzy sets is used to enhance the interpretability of the initial model. A genetic algorithm is used to improve the precision of the model. The process continues iteratively until the stop criteria are satisfied. The proposed approach is applied to the Iris benchmark classification problem, and the results show its validity.
出处
《电子学报》
EI
CAS
CSCD
北大核心
2006年第1期83-88,共6页
Acta Electronica Sinica
基金
国家自然科学基金(No.60332020)
南京理工大学科研发展基金资助计划项目(2005)
关键词
模糊分类系统
模糊聚类
遗传算法
解释性
精确性
fuzzy classification system
fuzzy clustering
genetic algorithm
interpretability
precision
