Optimal granule level selection: A granule description accuracy viewpoint

摘要

Granule description has become one of the hot research topics in granular computing (GrC). Rough set theory (RST), as an important research technique for granular computing, can describe any target granule (any subset of a universe of discourse) by the lower and upper approximations. But, there is no measure to evaluate the quality of granule description in RST. Moreover, one can acquire different granule descriptions by decomposing a multi-scale information table into different single-scale information tables. Then, it is important to find the most appropriate single-scale information table for meeting specific granule description accuracy requirement. Inspired by the above problem, this paper is to discuss optimal granule level selection based on the granule description accuracy. First of all, a new granule description accuracy is defined by combining GrC and RST. After that, optimal granule level selection is investigated in a multi-scale information table subject to preserving granule description accuracies for a target granule and a group of target granules, respectively. Specially, for the case of a group of target granules, we put forward optimal granule level selection methods based on three different criteria, commonly used by people in daily life. In addition, considering that the data in real-life will be updated as time goes by, we discuss the impact on the optimal granule level when new objects are added gradually. Finally, the time complexity of the proposed algorithms is analyzed, the reasonability of setting the parameters is explained, some numerical experiments are conducted to show the effectiveness of our methods, and a comparison of our algorithms and the existing ones is made. (C) 2019 Elsevier Inc. All rights reserved.