论文部分内容阅读
启发式方法是模糊粗糙集属性约简的有效手段之一,大多基于贪心策略的启发式方法以串行方式运行,这限制了属性约简的规模。为了解决该问题,提出一种基于并行计算的模糊粗糙集属性约简算法,该算法通过调用多个处理器来协同寻找对分类任务最具影响的最小属性子集,从而扩大了可约简数据的规模。在10组UCI数据集上的实验结果表明,基于并行计算的模糊粗糙集属性约简方法相比传统的启发式属性约简方法而言,在时间开销上能有很大程度的缩减。
Heuristic method is one of the effective methods to reduce attribute reduction of rough set. Most heuristic method based on greedy strategy runs in serial mode, which limits the size of attribute reduction. In order to solve this problem, this paper proposes a fuzzy rough set attribute reduction algorithm based on parallel computing. This algorithm calls for multiple processors to collaborate to find the smallest subset of attributes most influential to classification tasks, The size of Experimental results on 10 sets of UCI datasets show that the fuzzy rough set attribute reduction method based on parallel computing can reduce the time overhead greatly compared with the traditional heuristic attribute reduction method.