Single-parameter decision-theoretic rough set

Abstract Decision-theoretic rough sets (DTRSs), which can be considered as generalized rough set models produced by Bayesian risk minimum and three-way decisions (3WD) theories, have achieved fruitful results in risk decision-making problems. Nevertheless, the parameter determination of decision-theoretic rough sets is a challenging problem in practical applications, which narrows the generalization and development of these models. In this paper, a methodology to determine the parameters for DTRS and 3WD is proposed to improve their practicability. First, a data-driven loss function matrix is introduced based on the significance and the probability of the sample. Subsequently, a generalized rough set model named single-parameter decision-theoretic rough set (SPDTRS) is put forward based on the proposed loss function matrix. The main feature of the proposed model is that there is only one parameter that should be preset rather than the two or six parameters in the traditional DTRS models. Finally, some experiments on the University of California Irvine (UCI) and Knowledge Extraction based on Evolutionary Learning (KEEL) data sets are conducted to illustrate the effectiveness of the proposed methodology.