Andriichuk O. A method for defining semantic similarity of objects in knowledge bases of expert decision support systems

The dissertation is dedicated to development of a method for defining semantic similarity of objects in knowledge bases of expert decision support systems. The method provides an opportunity to improve the quality of recommendations for decision-makers through increasing the adequacy of models of weakly structured domains. The author suggests a method, which, in contrast, to other existing methods, does not require a training sample of object formulations, and is based on expert data utilization. In order to verify the relevance of the developed method, a simulation model of expert estimation has been developed, and a modeling complex has been created on its basis. The modeling complex provides an opportunity to test expert decision support methods without conducting costly expert examinations involving real experts. To facilitate expert estimation within semantic similarity definition method, a special software toolkit has been developed and tested. In contrast to other existing software toolkits, it allows the expert to use different scales for every pair comparison, and gradually increase his/her estimates, thus, preventing information distortion during expert data input.