Forecasting the fund of time for performance of works in hybrid projects using machine training technologies

Abstract

The aim of the work is to substantiate the approach to forecasting the time fund for work in hybrid projects, taking into account the changing nature and climatic components of the design environment based on the use of neural networks. The neural network architecture involves the use of a multilayer perceptron, teacher training, and the method of backpropagation. It is based on an algorithm that minimizes the prediction error by propagating error signals from the network outputs to its inputs, in the direction opposite to the direct propagation of signals. Based on the prepared initial data, the training of an artificial neural network was performed, which ensured the creation of an artificial neural network that is able to predict the duration of naturally allowed time to perform work in a software environment written in Python. Studies based on neural network training show that when the number of epochs increases to more than 25,000, the error does not exceed 4.8%. The obtained results indicate that the use of the proposed architecture of the artificial neural network gives a fairly accurate forecast and this is the basis for making quality management decisions on planning the content and timing of work in hybrid projects.