Decision-Theoretic Rough Set: A Fusion Strategy

Decision-theoretic rough set is a popular topic. However, such single-granulation rough set model is not able to handle complex information well, such as multi-source, multi-scale and high dimensions data. Therefore, the fusion of the ideas of Bayesian decision and multi-granulation may be an appealing issue. In this article, a novel rough set model based on multi-granularity decision theory is proposed. The discussed rough set model not only overcomes the shortcomings of optimistic and pessimistic rough sets, but also gains high approximation quality and low decision cost at the same time with a satisfactory threshold. In information granule reduction, heuristic and genetic algorithms are used to compute reducts based on three different criteria, respectively. The experimental results express that decision preservation based reduction may not suitable in such rough set models. Moreover, we also reveal that decision monotony and cost minimum based reductions are able to be popular research topics in rough set model of multi-granulation decision theory.