Abstract: S-probability rough sets and dominance-based rough sets are both important extensions of rough set theory.However, they both have their own limitations. Based on the advantages of the above two models, a two-direction singular dominance-based probability rough set model is proposed. The new model not only considers the dynamic statistical characteristics of the system but also takes into account the effective processing of actual preference information system. The model has certain ability to resist noise interference. At the same time, it has higher classification accuracy and classification plausibility, which is suitable for dealing with dynamic multi-attribute decision-making problems with preference information.

Key words: dominance-based rough set, S-dominance-based probability rough set, probability approximation accuracy, fault tolerance