摘要
提出一种可同时构造多个精确性和解释性较好折中的高维模糊分类系统的设计方法。该方法由以下三步组成:(1)利用Simba算法进行特征变量选择;(2)采用模糊聚类算法辨识初始的模糊模型;(3)利用匹茨堡型实数编码的多目标遗传算法对初始模糊模型的结构和参数进行学习。基于NSGA-Ⅱ算法的目标函数同时考虑模型的精确性和解释性;为提高模型的解释性,在多目标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等问题进行分类,仿真结果验证了方法的有效性。
A novel approach to construct accurate and interpretable high-dimensional fuzzy classification system was proposed. The approach is composed of three phases: (1) feature selection is accomplished by the Simba algorithm; (2) the initial fuzzy system is identified using the fuzzy clustering algorithm; (3) the structure and parameters of the fuzzy system axe optimized by the Pittsburgh-style real-coded genetic algorithm. The three-objective function based on NSGA-Ⅱ algorithm combines the interpretability index and the precision index. In order to improve interpretability of the fuzzy system, the similarity-driven rule base simplification techniques were used to reduce the fuzzy system, The proposed approach was applied to two benchmark problems, and the results show its validity.
出处
《系统仿真学报》
EI
CAS
CSCD
北大核心
2007年第1期210-215,共6页
Journal of System Simulation
基金
国家自然科学基金资助项目(60474034)
关键词
模糊分类系统
特征变量选择
模糊聚类
PARETO最优解
遗传算法
解释性
fuzzy classification systems
feature selection
fuzzy clustering
Pareto optimal solution
genetic algorithm
interpretability