Research of intellectual management models based on classification procedures of uncertain data with established requirements of result reliability

Abstract

For a wide range of complex systems, tasks such as selection of control options for various technological processes, selection of performers for assigned tasks, and determi-nation of authorship are resolved through classification and diagnosis of incomplete data regarding states and conditions of operation. The relevant problems include forming ad-equate mathematical models of classification procedures and establishing their correct-ness, completeness, and reliability of results. This article focuses on investigating the properties and development of intellectual management models for complex systems un-der conditions of data uncertainty based on classification procedures using reduction methods and Cohen's kappa statistics. It is noted that the application of these methods ensures reliable resolution of classification tasks considering the assessment of the max-imum model dimensionality. Additionally, the possibilities of improving Hamming neural networks intended for data classification tasks in formats of fuzzy values and certainty factors CF(A) were explored. The features of the proposed enhanced mathematical model for fuzzy classification tasks based on a set of feature templates defining the classes of objects under analysis were identified. The article also discusses the peculiarities of the mathematical model of classifi-cation designed for the task of determining the authorship of Ukrainian-language works (UAW). The characteristics of the UAW task and its implementation based on a fuzzy classification model include the absence of requirements regarding the number of stages in the authorship determination procedure, the unnecessary formation of a unified classi-fication model for UAW tasks for any possible input works, and the absence of the need to transform template models when introducing new data or works into the model. The listed features of classification procedures are accounted for in the reduction and Cohen's kappa procedures outlined in the article. To implement and study classification tasks of complex system parameters under conditions of uncertain data, appropriate software was developed. The article presents the structure of the software complex for information technology management of per-former assignment/selection, as well as the task of determining authorship of Ukrainian-language works based on classification of sets of templates with certain fuzzy features. The software complex utilizes reduction and kappa statistics procedures.