摘要
提出一种可同时构造多个精确性和解释性较好折衷的高维模糊分类系统的设计方法.该方法首先利用Simba算法进行特征变量选择,然后采用模糊聚类算法辨识初始的模糊模型,最后利用Pareto协同进化算法对所获得的初始模糊模型进行结构和参数优化.其中,Pareto协同进化算法采用了一种新的基于非支配排序的多种群合作策略.为提高模型的解释性,在Pareto协同进化算法中利用基于相似性的模型简化方法对模型进行约简.利用该方法对Wine典型问题进行分类,仿真结果验证了方法的有效性.
A novel approach for constructing accurate and interpretable high-dimensional fuzzy classification systems is proposed. First, feature selection is accomplished by the Simba algorithm; secondly, the initial fuzzy system is identified using the fuzzy clustering algorithm; finally, the structure and parameters of the fuzzy system are optimized by the Pareto co-evolutionary algorithm. The Pareto co-evolutionary algorithm is calculated by a new non-dominated sorting method. In order to improve the interpretability of the fuzzy system, the similarity-driven rule-based simplification techniques are used to reduce the fuzzy system. The proposed approach has been applied to Wine benchmark problem, and the results show its validity.
出处
《东南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2008年第4期626-631,共6页
Journal of Southeast University:Natural Science Edition
基金
江苏省博士后科研资助计划资助项目(0702027B)
江苏省自然科学基金资助项目(BK2006202)
关键词
模糊分类系统
模糊聚类
PARETO解
协同进化算法
解释性
fuzzy classification system
fuzzy clustering
Pareto optimal solution
co-evolutionary algorithm
interpretability