An approach for incremental maintenance of approximations in set-valued ordered decision systems while updating criteria values

In multi-criteria decision analysis problems, attributes are criteria with preference-ordered scales and the decision classes are also preference-ordered. Dominance-based rough set approach is an extension of classic rough set approach, which taking into account the ordering properties of criteria. Set-valued decision system is a generalized model of single-valued decision system. Knowledge from decision systems will vary with time under the dynamic decision-making environment. Incremental learning is an effective tool to deal with learning tasks since it can make full use of previous knowledge. In this paper, properties for dynamic maintenance of approximations are analyzed when the criteria values in the set-valued ordered decision system evolve with time. Then, algorithms for updating approximations incrementally with the variation of criteria values are discussed. Furthermore, experiments are carried out on several data sets to verify the performance of the proposed algorithm.