Kuvaieva V. Models and methods for aggregation collective expert’s estimations in rank scales for network decision-making systems

The dissertation is devoted to the development of models, methods and information technology (IT) of collective expert evaluations aggregation in rank scales for network decision support systems. The work is based on the aggregation of individual preferences of the team members, which can be formed from of the members of the network community who act as experts. Preferences are expressed in rank scales. The method for preliminary processing of network collective expert estimation (NCEE) rank matrix is proposed, which is formulated as a consecutive two-dimensional reduction of the matrix, with the exception of non-significant alternatives and experts with an opinion that is significantly different from the collective one. This allowed to significantly increase the consistency of the collective expert assessment and ensure the coefficient of concordance W > 0.5. The method for a collective expert estimation fast computation based on the median of Kenny-Snell and Cook-Seiford is proposed. It is based on the median calculation problem conversion to a well known assignment problem, which made it possible to reduce the computational complexity of the problem to polynomial and carry out median calculations for the number of alternatives n <50 in real time scale The model of calculation of collective expert evaluation on the basis of the theory of social choice using the Schulze voting principle was investigated, thus reducing the time of calculations of NCEE to 1-2 s with the number of alternatives n <50, while preserving the accuracy of the calculation of the collective evaluation. The two-stage method of calculation of collective expert evaluation on the basis of the Markov model and the Copeland rule has been developed, which allowed to reduce the time spent on calculating NCEE by 10-20%. The method for increasing the security of collective expertise from falsifications, based on the application for the organization of expert examination of blockchain technology, is proposed. IT of NCEE based on the proposed models and methods has been developed to build network decision support systems. Keywords: information technology, network collective expert estimation, rank scale, aggregation models, expert estimation reliability.