摘要
针对权重粗糙集模型不能有效处理非平衡混合数据的问题,对权重论域上的各种类型变量进行分析并建立统一的模糊等价关系,提出混合数据上的权重模糊粗糙集模型,并利用该模型构造出带权模糊等价空间上的混合属性约简算法.混合属性约简算法产生的模糊软划分可以克服权重论域上离散硬划分产生的信息损失.在非平衡混合数据集上进行的实验结果表明,与基于权重粗糙集的算法相比,基于权重模糊粗糙集模型的属性约简算法的平均分类精度提高了11.9%.
In order to solve the problem that weighted rough sets model lacks a mechanism to deal with mixed and imbalanced data, a unified fuzzy equivalent relationship for analyzing different types of features in weighted domain is established, and a weighted fuzzy rough sets model is proposed to deal with mixed data. Furthermore, a hybrid attribute-reduction algorithm is constructed based on the weighted fuzzy rough sets model. Compared with the classical crisp partition, the hybrid algorithm can avoid information loss through fuzzy soft partition generated by the model. Experimental results on imbalanced and mixed data sets show that the proposed weighted fuzzy rough sets model can not only select fewer features than weighted rough sets model, but also improve the average classification performance of the reduced attribute set on learning methods by 11.9%.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2011年第10期43-47,共5页
Journal of Xi'an Jiaotong University
基金
国家高技术研究发展计划资助项目(2007AA010408)
关键词
粗糙集
混合数据
模糊等价关系
权重论域
属性约简
rough set
mixed data
fuzzy equivalent relationship
weighted domain
attribute reduction