关键词: 数据挖掘, 粗糙集, 属性约简, 并行算法

Abstract: By the research of data mining rough set reduct, the author propounded a data dividing thought based on the discernibility ability of i ndex. First the information system was divided into many small tables, then the final reduction was resolved from the small table reduction. This thought has pre digested the reduction of the bool function reduction. And based on this thought, a parallel algorithm of attribute reduction was designed. By using the simulate tool CPN Tools of Petri Nets, the load balance of the algorithm was analyzed. On the basis of the experimental results compared with the corresponding serial algorithm, it is shown that the parallel algorithm is more efficient for the large scale system with excessive objects.

Key words: data mining, rough set, attribute reduction, parallel algorithm