Development of a methodology for training artificial neural networks for intelligent decision support systems

Oleg Sova,Alexander Simonenko,Yevhen Kalashnikov,Yurii Zhuravskyi,Ruslan Zhyvotovskyi,Olha Salnikova,Oleksandr Zubov,Andrii Shyshatskyi,Artem Shulhin,Іgor Romanenko

doi:10.15587/1729-4061.2020.199469

Abstract

A method for training artificial neural networks for intelligent decision support systems has been developed. The method provides training not only of the synaptic weights of the artificial neural network, but also the type and parameters of the membership function, architecture and parameters of an individual network node. The architecture of artificial neural networks is trained if it is not possible to ensure the specified quality of functioning of artificial neural networks due to the training of parameters of an artificial neural network. The choice of architecture, type and parameters of the membership function takes into account the computing resources of the tool and the type and amount of information received at the input of the artificial neural network. The specified method allows the training of an individual network node and the combination of network nodes. The development of the proposed method is due to the need for training artificial neural networks for intelligent decision support systems, in order to process more information, with unambiguous decisions being made. This training method provides on average 10–18 % higher learning efficiency of artificial neural networks and does not accumulate errors during training. The specified method will allow training artificial neural networks, identifying effective measures to improve the functioning of artificial neural networks, increasing the efficiency of artificial neural networks through training the parameters and architecture of artificial neural networks. The method will allow reducing the use of computing resources of decision support systems, developing measures aimed at improving the efficiency of training artificial neural networks and increasing the efficiency of information processing in artificial neural networks

Highlights

Decision support systems (DSS) are actively used in all areas of human life
The creation of intelligent DSS was a natural continuation of the widespread use of classic DSS
The aim of the study is to develop a methodology for training artificial neural networks for intelligent decision support systems, which allows processing more information, with the uniqueness of decisions made

Summary

Introduction

Decision support systems (DSS) are actively used in all areas of human life. They received special distribution while processing large data sets, providing information support to the decision-making process by decision-makers.Nowadays, artificial intelligence methods are the basis of existing DSS [1,2,3,4,5,6,7,8,9,10].Mathematics and cybernetics – applied aspectsThe creation of intelligent DSS was a natural continuation of the widespread use of classic DSS. Decision support systems (DSS) are actively used in all areas of human life They received special distribution while processing large data sets, providing information support to the decision-making process by decision-makers. Intelligent DSS provide information support for all production processes and services of enterprises (organizations, institutions), including product design, manufacturing and marketing, financial and economic analysis, planning, personnel management, marketing, support for product creation (operation, repair) and prospective planning. These intelligent DSS have been widely used for specific military tasks, namely [1, 2]:

Objectives

Methods

Findings

Conclusion