摘要
为进一步提升模糊粗糙集的分类性能,提出随机模糊粗糙分类策略。在原始属性集合上利用随机多重采样的方法,得到一组由原始属性的子集构成的合集,在此基础上采取投票方法实现模糊粗糙分类结果的集成输出。实验结果表明,这种随机模糊粗糙分类器的性能优于传统的模糊粗糙集和随机森林方法,在利用约简时,分类性能能够进一步得到提升,为使用集成思想研究模糊粗糙集的分类机制提供了可行的解决途径。
To further improve the performance of fuzzy rough classifier,a random fuzzy rough classification strategy was proposed.A set of attributes subsets was obtained using random multiple samplings on raw attributes.The voting strategy was employed and the outputs of fuzzy rough classifiers were integrated.Experimental results show that the performances of the proposed random fuzzy rough classifier are superior to that of traditional fuzzy rough set and random forest approaches,the performances can be further improved when attribute reduction is executed.A reasonable solution is provided for ensemble fuzzy rough classification mechanism.
出处
《计算机工程与设计》
北大核心
2017年第10期2814-2819,共6页
Computer Engineering and Design
基金
国家自然科学基金项目(61572242
6150316
61502211)
江苏省高校哲学社会科学基金项目(2015SJD769)
中国博士后科学基金项目(2014M550293)
关键词
属性约简
集成学习
模糊粗糙集
启发式算法
随机模糊粗糙分类器
attribute reduction
ensemble learning
fuzzy rough set
heuristic a lg o r ithm
random fuzzy rough classifier