Search academic texts

Information × Registration Number 0219U005006, 0115U004676 , R & D reports Title Intelligent information technologies of diagnosing and automatic classification. popup.stage_title Head Subbotin Sergey Alexandrovich, Registration Date 10-04-2019 Organization Zaporizhzhia National Technical University popup.description2 The object of study is diagnostic and automatic classification processes. The purpose of the work is the development and research of intelligent information technologies for the construction of diagnostic models by precedents. In the development of intelligent information technologies for diagnosing and automatic classification in the work the analysis of known methods of constructing diagnostic models according to precedents has been carried out, methods and means of intelligent processing of large data sets, diagnostics and automatic classification, as well as inductive synthesis of models by precedents have been developed. The method of synthesis of neuro-fuzzy models, which includes the most important features and terms in the model, eliminates the dubbing of terms and features, provides a grouping of features. The proposed method allows to significantly accelerate the synthesis of neuro-fuzzy models, providing an acceptable accuracy and a higher level of data generalization, reduce complexity and redundancy, as well as enhance the interpretation of the neural model. The method of synthesis of radial-base neural networks, in contrast to the known methods, does not require a user's task with the number of clusters, there is no uncertainty in the choice of the number of neurons in the first layer and the choice of the initial values of the weights of the network, seeks to minimize the size of the network, characterized by acceptable training time, takes into account informativity the signs in the formation of the breakdown into clusters, thanks to the use of optimization of the network allows to receive non-redundant, contrast, interpretsble models, and also provides the opportunity news relearning previously built models. A set of metrics is defined that allow to quantify the individual properties of the material and automate the analysis of niche documents. Using the proposed metrics allows you to automate the analysis and comparison of scientific documents. The results of a scientific work are recommended for use in solving diagnostic tasks, pattern recognition on the features, etc. Product Description popup.authors Благодарьов Олексій Юрійович Бондаренко Валерій Олегович Зайко Тетяна Анатоліївна Каврін Дмитро Анатолійович Колпакова Тетяна Олексіївна Олійник Андрій Олександрович Пришляк Михайло Юрійович Субботін Сергій Олександрович popup.nrat_date 2020-04-02 Close