NEURAL NETWORK MODEL FOR AVERAGE ANNUAL NET EARNINGS DETERMINATION IN EU ECONOMIES

Abstract

In macroeconomic modeling, the application of artificial neural networks (ANN) for forecasting key macroeconomic variables is present to a large extent. Precise forecasts of economic conditions are of crucial importance for the rational planning of overall economic efficiency in order to reduce the uncertainty of future economic trends. The purpose of constructing economic models is to determine the influence of changes in the value of certain economic variables on the model defined output variable. 
 Starting from earnings as a key concept in economics, the aim of the research is to produce a forecast model for determining the average annual net earnings, depending on the movement of key economic variables, i.e. from measures of economic performance. Based on a series of data relating to the period from 2005 to 2016, for the countries of the European Union, a high-reliability forecast model was developed using the most widely utilized type of ANN – a Multilayer Perceptron with a backpropagation algorithm. Specifically, the forecast model built to determine the earnings showed an average deviation of the real from the forecast size of up to 13%. Understanding and using such and similar models is crucial for policymakers when creating economic policies.