摘要
模糊粗糙集结合了粗糙集和模糊集的优势,使得可以直接处理连续值属性,成功应用于诸多领域.然而,在信息爆炸的时代,现实世界的数据绝大多数都在不断发展变化,传统的模糊粗糙集已不再适用于动态变化的数据.针对动态增长或变化的数据,将并行约简理论引入至模糊粗糙集中,结合信息论,提出信息论意义下的模糊粗糙集并行约简概念,并利用互信息的概念提出了基于互信息的模糊粗糙集并行约简算法.实验结果表明本文所提算法与传统的模糊粗糙集约简方法保持相近的分类准确率,但避免了繁琐的重新训练过程,大大提高了学习速度,因此,提高了分类的可扩展性和适应性.
Fuzzy rough sets combine the advantages of rough sets and fuzzy sets. It can handle continuous attributes directly, and has been successfully applied in many fields. However, in the era of information explosion, the majority of data in the real world are always developing and changing. The traditional fuzzy rough sets are no longer applicable to dynamic data. To deal with the dynamically growing or changing data sets, parallel reduction is introduced into fuzzy-rough sets. Combined with the information theory, fuzzyrough parallel reduction in information view is proposed, and the corresponding algorithm with the concept of mutual information is also presented. The experimental results show that the proposed approach can maintain almost the same classification accuracy with classical fuzzy rough sets,but it avoids the tedious re-training process so as to greatly improve the learning speed. Hence, the scalability and adaptability of classification can be increased.
出处
《小型微型计算机系统》
CSCD
北大核心
2015年第8期1847-1851,共5页
Journal of Chinese Computer Systems
基金
国家自然科学基金项目(61305094
61203240)资助
上海市教育委员会科研创新项目(12YZ140)资助
上海高校青年教师培养计划项目(sdl11003)资助
关键词
模糊粗糙集
并行约简
模糊决策表
互信息
fuzzy-rough sets
parallel reduction
fuzzy decision table
mutual information
