Gaussian Kernel Interval Type-2 Fuzzy Rough Reduction And Rules Extraction

With the fast development of computer technology and network information technology,the infomation and data obtained from various fields expand rapidly.They have high uncertainty,and the relationship between them is more complex(complex information system).Interval type-2 fuzzy rough set is a mathematical theory for dealing with imprecise and uncertain information which combines rough sets and interval type-2 fuzzy sets.The interval type-2 fuzzy set is introduced to improve the uncertainty of data,and it can analysis and deal with continuous data better.When studying the combination of fuzzy set and rough set,we need to focus on constructing fuzzy similarity relation reasonably.Gaussian kernel function has showed significant advantage in nonlinear division of numerical and fuzzy data which can produce better fuzzy relation in analyzing and dealing with fuzzy rough data.In this paper,we first study the concepts and theories of interval type-2 fuzzy sets,rough sets and interval two fuzzy rough sets.An attribute reduction algorithm for interval type-2 fuzzy rough sets based on Gaussian kernel is proposed.Type-2 fuzzy set theory is introduced to type-2 to fuzz the continuous attributes in fuzzy decision table.And Gaussian kernel function is introduced to construct fuzzy similarity relation better.Then the upper and lower approximation of type-2 fuzzy rough approximation space can be obtained,and attribute reduction of fuzzy decision table can be conducted.Then the fuzzy formal concept analysis theory is studied and the corresponding rule extraction algorithm is proposed.Fuzzy concept lattice rules are extracted based on the reduction results,and the classification rules can be obtained.At last,some experiments on UCI data sets are conducted in this paper.The experimental results show that the proposed method is feasible and effective.Finally,the algorithm was applied to the brain data,and relevant attribute reduction and rule extraction experiments were carried out.We can judge the artificial stimulation type by the activated degree of each brain region.