基于依赖度的区间集决策信息表属性约简

区间集决策信息表拓展了经典决策信息表,但其属性约简研究较少.针对区间集决策信息表存在的问题,采用模型正域及相关依赖度提出属性约简及其启发式约简算法.在区间集粗糙集模型中,定义关于决策分类的正域与依赖度,证明粒化单调性等性质.提出基于依赖度的属性约简,设计启发式约简算法.实例分析与数据实验表明,设计的基于依赖度的启发式约简算法是有效的,所得结果有利于依赖学习与特征优化.

Attribute reduction based on dependency degree for interval-set decision information tables

Interval-set decision information tables extend classical decision information tables. However, few studies have focused on its heuristic reduction. To solve the problem in interval-set decision information tables, this paper proposed attribute reduction and its heuristic reduction algorithm by adopting the model positive region and corresponding dependency degree. In the interval-set rough set model, it defined the positive region and dependency degree for decision classification, and proved the granulation monotonicity. It established attribute reduction based on the dependency degree and designed a heuristic reduction algorithm. The instance analysis and the experimental results show that the proposed attribute reduction algorithm is effective. The obtained results are useful for dependency learning and feature optimization.