Timochko O. Software Quality Evaluating Methods for Classifying Air Objects in an Automated Air Traffic Management System

The dissertation is devoted to the solution of the actual scientific problem related to the development of methods for evaluating the quality of software for the classification of air objects in an automated air traffic management system, which is essential for improving air traffic safety. The scientific novelty of the obtained results is that: for the first time, a method for verifying the software of a fuzzy logical classification system for aerial objects has been developed, that converts a database of fuzzy production rules into a workable model based on the use of fuzzy colored Petri nets within formal software verification methods; has received further development of the software classification method, which differs from the known complex processing of information from heterogeneous sources; formalization of software features based on binary relations and divisional procedures; a hierarchical four-tier structure of classification of air objects; features of the structure of the singleton knowledge base; analysis of the completeness and reliability of the input information to determine the superiority of the solution options; features of the mechanism of logical deduction, which has determined branches of decision-making about air objects and branches of decision-making on the basis of fuzzy logical conclusion with the formation of the vector of "trust" to the variants of decision; the method of estimating the quality of information for the classification of air objects has been improved, in which, unlike the known ones, the minimum distance of the available information from the reference is presented for the completeness criterion; prioritizing decision making using only the informative attributes inherent in each class when calculating. The main results of the research are recommended to be used in the development and testing of software for the classification of aerial objects, in particular, in the context of the processing of characteristic information coming from various sources.