STORAGE OF ARTIFICIAL NEURAL NETWORK OF BUSINESS MANAGEMENT

Abstract

This publication examines the problem of data storage modeling using artificial neural networks. Such a repository allows you to collect, store and analyze data, which contributes to making informed decisions and maintaining competitiveness. Using TensorFlow as the basis for the data warehouse provides additional possibilities for processing business management information data from various sources, including databases, Internet resources, sensors, and more. This data can be stored as tables or files and further processed to train the model. The feature of the proposed model is the presence of one hidden layer with 10 neurons and the use of the ReLU activation function. To improve the accuracy of the model, the MSE loss function and the Adam optimizer are used, which allows changing the network weights. After training, the model can evaluate the accuracy on the test data and make predictions for the future period. Inputting new data allows the model to make predictions that can be evaluated using different metrics, depending on the intended use.