The Fuzzy Inference Rule Extraction And Attribute Reduction Based On AFS Theory And Closeness Degrees

Today a lot of important information and knowledge are included in all kinds of database. What we really concern about is how to find useful information and benefit more from them for decision workers. AFS theory was proposed first by Liu Xiaodong in 1995. It is a new field of fuzzy mathematics and the membership functions and their fuzzy logic operations can be impersonally determined according to the original data. It embodies the abstract relationship of human conceptions using topological molecular lattices. In other theories the membership functions of fuzzy sets are defined artificially by human intuition.In this paper AFS theory and closeness degrees are combined to analysis attributes. A new fuzzy inference method is proposed. The examples show that the results are not only similar to human intuition but also objectively reflect the hided modes in original data. By the way we combine AFS theory and closeness degrees to analysis the relations among attributes in database and a new method of attributes reduction is given. By an example we can show that this method is precise and simple. Comparing to current research, this algorithm can be applied in database with variety data types such as integer real number, Boole value and sub-preference. So it is better than the method of coarseness sets in some aspects. Some attributes that can't be reduced by coarseness sets can be treated freely by this method.The methods of fuzzy inference rule and attribute reduction based on AFS theory and closeness degrees mentioned in this paper depend on AFS theory so they are more objective and uniform. The methods are favorable to data mining and easily operated by computers. We can serve the society with them in some ways.