The Methods Of Rule Extraction And Attribute Reduction Based On The Linguistic-Valued Formal Decision Context

With the rapid development of big data context,the information environment of people is also imprecise.As a tool that can effectively mine and analyze data,the concept lattice has been gradually applied to various fields,which is conducive to the processing of intelligent information.In real life,the information that people obtain is uncertain as well,and it is usually expressed by language which is close to human thinking.In order to reduce the loss of language information in the conversion process,this paper constructs the linguistic-valued layered concept lattice under the formal context with incomparable linguistic-valued information,and studies the methods of rule extraction and attribute reduction.The main research results are as follows:1?In order to reduce the loss of the linguistic-valued information conversion,the linguistic-valued formal context is given based on the linguistic-valued lattice implication algebra and concept lattice.By setting different linguistic-valued trust degrees,we put forward a linguistic-valued layered concept lattice for meeting the requirements of different experts at different levels.And we discuss the order relationship between the linguistic-valued layered concepts as well as properties.We construct the corresponding model of linguistic-valued layered concept lattice with the trust degree,and the effectiveness of the proposed method is validated by an example.Furthermore,we present the linguistic-valued weakly consistent formal decision context based on linguistic-valued formal context by introducing the concept of implication relation,which provides a new idea for reasoning and decision-making.2?Based on the linguistic-valued weakly consistent formal decision context,we deeply study the method of rule extraction,which provides a good scientific basis for uncertainty reasoning and decision-making with linguistic-valued information.By the subtle relationship between the linguistic-valued layered concept lattices which based on condition attributes and decision attributes with the trust degree,we extract the linguistic-valued rules under personalized requirements.Furthermore,to update and simplify the rules set so as to make the rules acquisition easier and the extracted linguistic-valued decision rules more compact.A linguistic-valued rules extraction algorithm with the trust degree is established,the example of evaluating the teaching level of teachers verifies the rationality of the method,and the superiority of the proposed method by comparing with other methods.3?To deal with the uncertainty information more efficient,we study the attribute reduction based on the linguistic-valued weakly consistent formal decision context.First,we eliminate the unnecessary attributes in the decision attribute set,and give the linguistic-valued irreducible formal decision context.Then for making the calculation time shorter and the storage space less,and improving the efficiency of handling problems,a method of attribute reduction based on the core concepts of discernibility matrix is constructed by using the mathematical tools of discernibility matrix and discernibility function.