Machine Learning Technique in Time Series Prediction of Gross Domestic Product

Abstract

Artificial intelligence is gaining ground the last years in many scientific sectors with the development of new machine learning techniques. In this research, a machine learning methodology is proposed in the Gross Domestic Product (GDP) time series forecasting. Artificial Neural Networks are implemented in order to develop forecasting models for predicting the Gross Domestic Product. A Feedforward Multilayer Perceptron (FFMLP) was implemented since it is considered as the most suitable in times series forecasting. In order to develop the optimal forecasting model, several network topologies were examined by testing different transfer functions and also different number of neurons in the hidden layers. The results have shown a very precise prediction accuracy regarding the levels of Gross Domestic Product. The proposed technique based on machine learning can be very helpful in public and financial management.