区间值决策表中基于相对知识粒度的属性约简

Attribute Reduction Based on Relative Knowledge Granularity in Interval-valued Decision Table

针对区间值决策表,采用区间相对知识粒度提出属性约简及其启发式约简算法。基于相似关系,定义关于决策分类的区间相对知识粒度,证明粒化单调性等性质;基于区间相对知识粒度,提出属性约简,挖掘区间属性内与外重要度的启发式信息,从而设计启发式约简算法并分析其时间复杂度;针对一致区间值决策表,证明区间相对知识粒度表示与代数表示的等价性;采用区间值决策表实例进行有效验证。对于区间值决策表,相关区间相对知识粒度及属性约简深化了知识学习与特征优化。

Aiming at interval-valued decision tables,attribute reduction and its heuristic reduction algorithm are proposed by adopting the interval relative knowledge granularity.Based on the similarity relationship,the interval relative knowledge granularity is defined for decision classification,and the granulation monotonicity is proved;by the interval relative knowledge granularity,attribute reduction is established,the interval attribute inner/outer significance is mined to become heuristic information,so a heuristic reduction algorithm is designed;the equivalence between the algebraic representation and relative knowledge granularity representation is proved for consistent interval-valued decision tables;finally,an example of interval-valued decision table is used for effective verification.Regarding interval-valued decision tables,the obtained interval relative knowledge granularity and attribute reduction deepen knowledge learning and feature optimization.