A Weighted Multi-Granulation Decision-Theoretic Approach To Multi-Source Decision Systems

Decision theoretic rough set is a typical generalization model of rough set, which has fault tolerance based on Bayes minimum decision risk. How to mine knowledge from the information collected from different sources is one of the focuses of current artificial intelligence. From a cognitive point of view, especially from the point of granulation, this paper studies decision theory of multi-source decision systems based on generalized multi-granulation and decision theoretic rough sets. It is because each granular structure is not equally important in practical issues. First of all, the method of granulation weight is proposed based on the internal uncertainty of systems and the external correlation between systems, namely the double weighted granulation (DGW) method. And then a weighted generalized multi-granulation decision-theoretic rough set (WGM-DTRS) model in multi-source decision systems is proposed. Finally, in order to verify the effectiveness of the (DGW) method, the approximation accuracy of decision classes under different weighted granulation methods is compared. The numerical results show that the proposed method is effective from the point of classification. Therefore, the WGM-DTRS model based on the DGW method is meaningful.