UDC 005.519.7
ENSEMBLES OF NEURAL NETWORKSWHEN FORECASTING VOLUME OF SALES IN TRADING NETWORK
V. O. Bryukhnova, student, group 744M, Ryazan; This email address is being protected from spambots. You need JavaScript enabled to view it.
N. I. Tsukanova, Ph.D. (Tech.), Associate Professor, department of calculating and applied mathematics, RSREU, Ryazan; This email address is being protected from spambots. You need JavaScript enabled to view it.
The questions of neural networks ensemble application to the decision of a problem of sales volumes forecasting in trading network are considered. Two methods are used to form the ensembles of neural networks, based on bugging and boosting methods. The aim of the work is to study the effectiveness of using neural network ensembles to solve the problem of sales volumes forecasting in trading network, to develop programs in MatLab environment that implement various algorithms for training ensembles of neural networks. The paper considers the following issues: the formulation of the problem to predict time series with the help of neural networks ensemble, model of neural networks ensemble in the form of individual neural
network models composition, learning and prediction algorithms based on the ensemble of neural networks in bugging and boosting methods, comparison of various algorithms in accuracy and stability of the forecast. The results of the research obtained with the help of the developed programs are given, on their basis recommendations on their use in forecasting sales volumes in sales network are given.
Key words: forecast, time series, regression, machine learning, model ensembles, running, boasting, learning error, generalization or testing error.