Research On Causal Discovery Method Based On Fuzzy Linguistic Formal Context

With the development of artificial intelligence technology,people have paid more and more attention to the credibility of artificial intelligence.Trustworthy AI needs to explore the interpretability and predictability of behaviors and results,that means to find causal relationships between attributes or features.But things often have fuzziness and uncertainty,it is very essential to explore the causal relationship under the uncertain environment.Linguistic value is one of the best ways to express uncertain information.It can imitate human thinking to achieve knowledge acquisition,which can conducive to the subsequent decision making or reasoning work.To solve the problem of inference of causality between linguistic value attributes under uncertain environment,this thesis studies the causal relationship between fuzzy attributes based on the fuzzy linguistic formal context and causal inference,and explores the causal inference between attributes under the fuzzy linguistic decision formal context,as well as the dominant granularity judgment method based on the causal certainty when the granularity of linguistic value attributes changes.The main research contents of this thesis are as follows:(1)Aiming at the problem of causality inference between attributes or features in uncertain environment,firstly,this thesis defines the attribute-induced fuzzy linguistic formal context and its properties,introduces the concept of associated attributes in attribute topology,obtains the coupling relationship between fuzzy attributes,and deduces the causality relationship between attributes according to the theory of causal conditions.In order to solve the problem of judging the causality between attributes under the fuzzy linguistic formal context,this thesis proposes the causality inference algorithm among attributes in the fuzzy linguistic formal context.According to the number of objects corresponding to linguistic value attributes and the weight between attributes,the order of deleting attributes is determined to realize the update of fuzzy linguistic formal context,and the causality between global attributes can be inferred.(2)In order to solve the problem of causality inference between attributes in the decision-making process,under the fuzzy linguistic decision formal context,the causal determining factor is defined according to the basic concept of causal inference,and the conditional attribute is intervened by the equivalence class induced by single attribute,which can realize to judge the degree of causality between the conditional attribute and the decision attribute.In an example,it can prove the validity of causal determining factors in judging the causal relationship between attributes under the fuzzy linguistic decision formal context.(3)Due to the different granularity of linguistic values of attributes,the idea of granular computing is used to determine whether the changes in the granularity of linguistic values of attributes can affect the causal relationship between attributes by setting different thresholds of causal factors,which can realize comparative analysis of causal relationship degree between attributes in a multi-granularity environment.Then,based on the causal determination degree between attributes,a judgment method of multi-granularity fuzzy linguistic decision formal context is proposed,and its rationality can be verified by an example.