Experience of Multi-Layer Neural Network Application for Empirical Regularities Identification

Abstract

In this article, the recognition method is applied to the comparison of industrial enterprises. At the same time, we had to consider the methods of taking into account non-formalized factors. The method can be used for objective ranking of enterprises. While researching the topic, a neural expert system was developed, and it was used by specialists of the regional government to predict the dynamics of industry.

In view of the complexity of relations, conflicting non-equality systems were arising. We were resolving them by means of committee method.

The effect of managing the situation by varying the values of factors by tracking them on a computer model, made it possible to adjust the model as a means of parallelizing the committee constructions. They act directly in multi-layer neural networks.