JEL Classifications: R11, R58
USE OF ARTIFICIAL NEURAL NETWORKS TO PREDICT TERRITORIAL ECONOMIC INDICATORS Author: Gocha Ugulava
| Published: 2019-12-27
| Pages: 143 - 146
For Citation: Ugulava, G. (2019). Use of Artificial Neural Networks to Predict Territorial Economic Indicators. Globalization And Business, #8, pp. 143-146
Modern economic science is unthinkable without predict- ing and planning the prospects for economic life development. There are many different mathematical and statistical tools in the arsenal of scientists as well as practitioners and econo- mists today in purpose of forecasting. To date, one of the most prominent effective tools for data analytics is artificial neural networks. Artificial Neural Network - is a mathematical mod- el created in the likeness of a human neural network, and its software and hardware implementation. We carried out mod- eling and forecasting of regional economic indicators using the artificial neural network of the three-layer perceptron archi- tecture. The network architecture and neuron settings were automatically formatted through the programming language R and its package - Neuralnet. During the forecasting phase, the data vectors were presented as data frame in five input param- eters (DFI, FAI, EMP, BT, CPI), according to the neural network forecast of the regional gross domestic product (RGDP_NN) was calculated. All data are from the Imereti region and are taken from official GeoStat sources. Forecasting was done at the same time scale (2006-2017) to enable us to compare the predicted values with the actual ones to verify the level of fore- cast accuracy. We also tested the results of the neural network in another way - compared to the predicted values using mul- tiple linear regression on the same data. The accuracy of the predicted values calculated by the neural network was quite high, which was not declining but slightly ahead of the accura- cy coefficients of the predicted values obtained through linear regression. Also, the predictive values calculated by the neu- ral network with high adequacy and accuracy were compared with actual, existing ones.
Presented material shows that the use of artificial neural networks for the prediction of territorial economic indicators is reasonable and justified. Their role in analyzing and predict- ing indicators that are characterized by non-stationarity, dy- namism, lack of a definite trend, periodicity, nonlinear struc- ture is especially increased. It is therefore advisable to apply this method in regional economic studies, in predicting terri- torial development plans, strategies, targets and indicators.
REGIONAL ECONOMICS, FORECASTING, ARTIFICIAL NEURAL NETWORKS, MATHEMATICAL MODELING, TERRITORIAL ECONOMIC INDICATORS
- Raul Rojas (Author), J. Feldman (1996). Neural Networks: A Systematic Introduction. Springer, 1 edition
- David Kriesel (2007). A Brief Introduction to Neural Networks. available at http://www.dkriesel.com
- Stuart Russell, Peter Norvig (2013). Artificial Intelligence: Pearson New International Edition: A Modern Approach. Pearson, 3 edition
- Haykin Simon (2018). Neural networks. Full course. Williams
- Stepanov M. F. (2010) - Artificial neural networks in planning and control tasks. Saratov, p. 5
- National Statistics Office of Georgia - www.geostat.ge