Attribute reduction is a key step in processing large-scale datasets.Compared with the tra-ditional Neighborhood Rough Set(NRS),the Granular Ball Neighborhood Rough Set(GBNRS)can significantly improve the performance of attribute reduction.However,the current GBNRS attribute reduction algorithms generate too many unnecessary granular balls,which greatly reduces the effi-ciency of the algorithm.This paper first defines a new granular ball quality metric to control the gen-eration of an adaptive number of granular balls.Then,it partitions the sample set using granular balls,placing sample points of different categories into corresponding granular balls.Finally,forward attri-bute reduction is performed based on the number of positive region samples within the granular balls under different attribute sets.To verify the effectiveness of the algorithm,comparative experiments were conducted with other NRS attribute reduction algorithms on 12 real datasets.The experimental results show that our proposed algorithm has higher accuracy and faster operational efficiency.
维普
万方数据