USING THE ANALYTIC HIERARCHY PROCESS WITH FUZZY LOGIC ELEMENTS TO OPTIMIZE THE DATABASE STRUCTURE

Abstract

Listen

Translate article

Context. Informational systems are very common and use databases to store information that users need. Many different data models can be used but the relational model is still relevant. The last decade show tendency of using distributed databases while working with relational data model and this approach requires a specially designed module to synchronize data of all separate databases. Considering optimizing the database structure, researchers didn’t pay much attention to the potential of users’ SQL-queries history. The optimal structure of all the distributed nodes could reduce the necessity of synchronization while the data access speed and its actuality would remain stable. The object of the research is the process of optimizing the structure of the distributed database of corporate information systems, which are based on the relational database’s model. Objective. The research aims at improving the accuracy of the data representation marker’s value on the distributed corporate information system’s (DCIS) node, obtained using the analytic hierarchy process by applying the fuzzy logic elements while processing the alternatives’ global priority vector. Method. The research’s authors in the set of their previous works emphasize the potential of using the collected history of users’ SQL queries. Firstly presented technology of users’ queries parsing. Then, the idea of using the multidimensional database for analyzing users’ queries by slices of workstation type, application, user, and his/her position was considered. Finally, the authors gave the full-scaled mathematical model for formalizing database and query models, and criteria of database structure’s optimality.The current research continues the given sequence and tries to increase the efficiency of the decision support system, by introducing elements of fuzzy logic to the analytic hierarchy process algorithm. The approach’s main idea is in presenting the global priorities vector in the form of a series of fuzzy sets of one variable with subsequent transformation to the exact value. This approach made it possible to maintain the accuracy of the obtained result while decreasing the number of solution alternatives. For new tuples added to the database’s tables after all calculations had been performed, the problem was formalized. After obtaining the probability of a tuple belonging to the class “needed” and performing the normalization of the value, it is taken as the level of the representation marker. Accordingly, the data is loaded onto the node if this value is greater than the optimal level of the representation marker for the DCIS node. Results. After calculating and obtaining the alternatives global priorities’ vector in order to improve the accuracy of the obtained result, the apparatus of fuzzy sets was used. The obtained vector of global priorities was presented as a vector of fuzzy digits for the data representation marker with subsequent transformation to the exact value. This approach made it possible to maintain the accuracy of the obtained result while decreasing the number of solution alternatives. Conclusions. While working on the research, the concept of a data representation marker on the DCIS node for the elements of the SQL query model was introduced. An aggregation function has been developed that allows determining the level of need for attributes and tuples in the database’s relation for the DCIS node based on the statistics of SQL queries. A model of the dependence of the database structure’s optimality criteria on the value of the data representation marker is built. Received further development method of analytic hierarchy process. The initialization of the alternatives’ pairwise comparisons matrix can be performed automatically according to the obtained mathematical models. Representation of the obtained result in the form of the vector of fuzzy numbers with the reduction to the exact value allows increasing the accuracy of the obtained results.