Research On Fuzzy Query Based On Fuzzy Cluster And Strategy Of Query Relaxation

Classic related database is based on Boolean logic, it means Classic related database can only represent and store precise and deterministic information. It also means that users can only inquiry accurately; however, a large number of users’ inquiries is vague and imprecise, therefore the study of fuzzy inquiry based on fuzzy clustering analysis is needed. However, either the intersection between the results of fuzzy query conditions and the results of exact condition is empty or error expression of the users may lead to failure query. This requires that the database be able to relax constraints on the initial query conditions to some extent, to get similar results with the initial queries.To solve these problems, this paper studies fuzzy query and failure queries deeply, and solutions are given respectively. This paper puts the method of using fuzzy clustering analysis to complete the fuzzy query in the relational database, which is based on the fuzzy logic theory, to get rid of the defects of the SQL which can only be used on accurate query. The method is applied as follows: The attribute values of the query are computed by the fuzzy clustering analysis, and then be dynamic classified by using the Prim algorithm, next, the data mapped to the corresponding fuzzy sets by using the mapping function, which means the clustering of data is applied by the characteristics of the query data itself. By the above efforts, we can get more reliable and general results of the query which could avoid the subjectivity of the fuzzy query by defining of membership functions artificially. Experimental results show that the use of fuzzy clustering analysis for fuzzy queries have got more ideal query results while avoiding the subjectivity.Secondly, in terms of the empty queries, this paper gives solutions in turn to solve fuzzy condition relaxation algorithm and precise query relaxation algorithm. For the fuzzy condition relaxation algorithm, this paper presents the ideas of rewrite the mapping functions to achieve a query relaxation, thus we can get more clusters in rewriting the mapping functions, in order to get more return results. For the algorithm of precise query, this paper proposed that to access the weight of the property firstly, and then to relax the property constraint based on the weight from small to large in turn. Relaxation is divided into non-specific numerical attributes and numerical attributes, for non-specific numerical attributes, the paper takes a coarse-grained query relaxation methods, namely the removal of the property directly; For numerical attributes, the paper puts forward the histogram-based query relaxation techniques, results from recall and precision demonstrated the relaxation method proposed by this paper is effective.