The Use of Artificial Neural Network for Estimating the Romanian Economic Sentiment Indicator

Abstract:

The present paper presents the results of a research that had us subject the simulation of the Romanian Economic Sentiment Indicator (ESI-mentioned by The Economist and provided by the European Commission – Economic and Financial Affairs website) using Artificial Neural Network (ANN). Imposing a smaller difference between the real data and the simulated data smaller than 5%, the researchers used a feed forward artificial neural network and a back propagation algorithm for the training and preparation of future use of the ANN. After a very good training result the ANN was tested three times using different data sets. All the tests were a success considering the initial conditions and the difference between the real ESI and the simulated ESI. In order to determine the future ANN use for the ESI forecasting, the research was extended by using new input data for the ESI simulations. These new data (these were not available at the time of ANN training) were used to forecasting the ESI trend in order to compare it with the real trend. The results of trends comparison was a difference with values between [-4.92; 5.16] %. So even with new data the ANN was able to offer forecasting with accuracy smaller than 5%. The training and use of ANN was considered a success, but the authors considered that the research
can be extended to other countries ESI or even to the European zone.