摘要
针对现有增量式属性约简算法中存在的约简传承性差以及不完备现象,提出基于标记可辨识矩阵的增量式属性约简算法.本文首先定义了标记函数,对样本之间的可辨识性进行分类,并将之引入一个新的可辨识矩阵,在新增样本时,结合标记信息可以快速识别可辨识矩阵元素集的异动,获得强传承性的约简超集,在此基础上,设计与标记可辨识矩阵匹配的必要矩阵,用以快速判断并删除冗余属性,确保约简的完备性.理论分析以及实验测试表明,本算法具有约简传承性强,约简集完备等特点,具有较强的实用性.
In order to improve the inheritance rate of reducts and obtain the complete reducts, an incremental algorithm for attribute reduction based on labeled discernibility matrix is proposed. The label function is first defined to classify the discernibility relationships of all object pairs. The labeled discernibility matrix is then proposed to find out the changed elements and compute a supper set of reducts quickly when a new object is added. At the same time, a necessary matrix corresponding to the labeled discernibility matrix is presented. It is used to delete the redundant attributes for a complete reduct. Theoretical analysis and experimental results show that the reducts calculated by the proposed algorithm are complete and have the characteristic of high inheritance rate.
出处
《自动化学报》
EI
CSCD
北大核心
2014年第3期397-404,共8页
Acta Automatica Sinica
基金
国家自然科学基金(61025015
61273159
61321003)
国家科技支撑计划(2012BAF03B05)资助~~
关键词
标记可辨识矩阵
必要矩阵
增量式约简
约简传承性
Labeled discernibility matrix
necessary matrix
incremental reduction
inheritance rate of reduct
