MACHINE LEARNING IN MANAGING THE PRODUCTION OF ENTOMOPHAGES

Abstract

The article is devoted to the issue of creating a multilayer neural network of direct signal propagation for intelligent decision support on the quality of entomological products in the production of entomophages. The relevance of the chosen direction of research is determined. The purpose of the study was to develop a neural network to solve the problem of classifying the quality of entomological products in the production of entomophages. The object of the study was the process of classifying the quality indicators of the Ephestia kuehniella in the production of the entomophages Habrobracon hebetor. Research methods – neural network and heuristic approaches, computer modeling. A three-layer forward signal propagation neural network was developed, which classifies the quality of Ephestia kuehniella in the production of the entomophage Habrobracon hebetor. The parameters of the input layer are the indicators of the quality of the Ephestia kuehniella - the mass of caterpillars of an older age, the parameters of the output layer - the quality class. The number of hidden layer neurons is calculated heuristically. To avoid retraining the network, training, control and test samples were formed. The network was trained using the Artificial Neural Network Scilab software package using the Levernberg-Marquardt algorithm. The average error of approximation of the learning results was (0.07-0.08) %, which indicates high classification accuracy. Research results make it possible to reduce the influence of the human factor in decision-making processes in the production of entomophages, to structure data on product quality.