Rules Acquisition Of Multi-granularity Decision Information System Based On Fusion Method

The application of multi-granularity rough sets is more and more extensive at present.It describes problems from different perspectives by constructing multiple binary relations,making the analysis of problems more thorough and representative.Attribute importance,sorting method and rule acquisition method are important researching topics of multi-granularity decision information systems.In order to effectively overcome the problem that objects with little difference are divided into different classes,a new method of data preprocessing is given.On this basis,the existing attribute importance measure is improved,and the concept of conditional attribute subset's weight and conditional attribute's weight under multi-granularity framework is proposed.At the same time,based on evidence theory,a sorting rule using weighted quasi-probability function is proposed,and this method can also be used for the sorting problem of multivariate objects.Then,in the multi-granularity framework,the concept of decision-making ability of a conditional attribute is proposed.Based on the Murphy evidence fusion method and the decision-making ability,for each object,a fusion method of decision rules in conditional attribute subset is given.Then the mass functions under different conditional attribute subset is fused by using the conditional attribute subset's weight and additive and multiplicative method,then the quasi-probability function of the fused possible decision rules is constructed.Based on this function,the fused decision acquisition method is given.It is generalized to the fusion method of mass function of system-based object subset in the whole universe of discourse,and the decision value and confidence degree of an object set can be obtained.Finally,the fusion method of decision rules for two multi-granularity decision information systems under the same universe is discussed,the overall influence of two decision rule on decision-making is defined,and a new method applicable to the synthesis of conflicting decision rule is given.And the effectiveness of the method is demonstrated by examples.