THE APPLICATION OF THE NEURAL NETWORKS FOR FORECASTING THE IMPORT OF GOODS FROM THE COUNTRIES OF THE EUROPEAN UNION

Abstract

In the article the topicality of using neural networks to the observe of the international trade, in particular, for modeling the main tendencies in changing the indicators of international trade was substantiated. The advantages of using the neural networks over the traditional methods in order to forecasting economic processes were described.

A neural network for forecasting the import volumes of goods of the countries of the European Union was constructed. The model was made by means of the form Statistics/Neural networks of the software package STATISTICA 10 and statistical data of import volumes of goods in million euro of EU countries during each month from January 2002 to September 2017.

The analysis of dynamics of import volumes of goods of EU countries during the investigated period was realized. Based on this analysis it is clear that over time the imports volume of goods of EU countries is growing and followed by a growing linear trend, as well as a certain periodicity. The expediency of applying the theory of time series to the observe of this process was described. The periodicity of the time series was determined by means of Fourier spectral analysis.

The constructed neural network forecasting model is a fairly correct representation of real statistics data. The adequacy of the constructed model was confirmed on the basis of cross-check (error 5.7%) and analysis of the model residues.

Using the received neural network model, predicted values for import volumes of goods of EU countries for future periods were calculated.