基于粗糙集理论的多标记数据互补决策约简加速算法

互补决策约简是一种多标记数据属性约简方法,当数据规模较大时,其启发式算法的计算耗时较大。基于粗糙集理论,对互补决策约简启发式算法的加速算法进行了研究。当粒度由粗变细时,在逐步去掉正域的数据集上,首先研究互补决策约简中属性外部重要度的保序性质;基于此,通过逐步缩小数据规模来降低计算约简的耗时,提出了互补决策约简加速算法。加速算法不仅减少了属性约简的计算时间,而且能够保持原始算法的约简结果。

Rough Set Theory Based Accelerated Algorithm for Complementary Decision Reduct of Multi-label Data

Complementary decision reduct is an effective attribute reduction approach for multi-label data. However, the corresponding heuristic algorithm is computationally time-consuming for large data sets. This paper proposes an accelerated heuristic algorithm of complementary decision reduct based on rough set theory. First, the rank preservation of outer significance measure of attribute in complementary decision reduct is studied on a dataset of positive region that is gradually removed when the granulation changes from coarse to fine. Then, an accelerated heuristic algorithm is proposed which can decrease the time-consuming by gradually shrinking the data scale. The accelerated algorithm not only speeds up the process of attribute reduction, but also preserves the reduction results of the original algorithm.