Research of possibilities of using neural networks in the decision support system

Abstract

The possibility of using neural networks in automated control systems for multi-assortment dairy production was considered. In the automatic control theory, many methods were developed that allow to optimize systems in terms of one or another quality criteria, provided that the number of restrictions is fulfilled, but mathematical tools used in the traditional automatic control methods, are not always able to fully ensure satisfactory results under a limited number of input data. Using neural networks allows to perform control of acceptable quality (not necessarily optimal) under uncertainty at a relatively low level of resources spent. During the research, the structure and learning algorithm of the neural network for the decision support system concerning the dairy plant assortment forecasting for the current day was determined. In the intelligent technology environment STATISTICA NeuralNetwork based on the model obtained, the sensitivity analysis of fuzzy neural network output to a change in the input stream was carried out. Using the neural network allows to take into account nonlinear dependences in problems of forecasting profitable dairy plant assortment, which is important for effective management under uncertainty. The research conducted are needed to develop the automated control system for multi-assortment dairy production.