Investigation of Hamming Network Procedures for Controlling Service Systems with Smprecisely Defined and Natural Language Data

Abstract

Purpose. Models and methods, as well as software tools for the tasks of planning the flow of orders of service systems, or service systems (S&S), are quite widespread. The task of developing processes for classification and management of S&S based on the associative memory model of the Hamming neural network (HNN) with imprecisely defined data characteristics is relevant today, and has theoretical and practical significance. The main purpose of the work is to develop and study mathematical models of Hamming network procedures for S&S with imprecisely defined and natural language data characteristics, comparative analysis of fuzzy set models and CF confidence coefficients. Methodology. The paper uses a modification of the Hamming neural network procedures and numerical experimental studies of the comparative possibilities of using fuzzy sets μX (X → [0; 1]) as models of primary data, as well as expert confidence indicators, confidence factors CF(A) from the set [–1; +1]. Findings. The formation and study of improved models of Hamming neural networks intended for classification procedures in S&S with imprecisely defined and natural language data characteristics is carried out. Originality. For the first time, the comparative possibilities of using fuzzy values (NVs) and CFs as models for representing the properties of incomplete and imprecisely defined data, as well as data in natural language form, are investigated for the tasks of classification and management of S&S. At the same time, the advantages of the CF confidence factor model are established and appropriate procedures for classifying and managing S&S are formed. Practical value. The models and procedures for classifying the properties of multi-parameter S&S objects based on modified Hamming neural networks developed in the article allow to effectively solve a wide range of tasks in the field of S&S management under uncertainty and incompleteness of primary data.