Key points are not available for this paper at this time.
In large-scale decision systems with high dimensions, constructing an efficient feature selection method via an uncertainty measure, has become a critical problem in fuzzy rough sets (FRS). However, the uncertainty method constructed through FRS for feature selection has the following limitations. 1) The composition of the uncertainty caused by fuzzy distance and similarity is neglected, which can not precisely evaluate the uncertainty. 2) The method of measuring uncertainty is to select all the sample for establishing a fuzzy similarity matrix, which leads to substantial time consumption. 3) The efficiency of selecting import features in a nonbatch manner is relatively low. Driven by this, both granular-ball (GB) computing and three-way approximations (TWA) are integrated to guide an uncertainty measure named multigranularity fuzzy entropy (MGFE), which is based on fuzzy distance and similarity, to improve the efficiency of feature selection. The MGFE is primarily recommended for measuring the uncertainty in multigranularity spaces. Therewith, the GB and TWA computing are integrated to compress the sample space to select representative sample. In addition, the MGFE is employed to assess the significance of the features in the representative sample space. Aided by the TWA, an efficient filter-wrapper feature selection with a three-way accelerator is successfully developed. Finally, related experiments illustrate the advancement of our proposed feature selection.s
Xia et al. (Wed,) studied this question.