国际经济与管理科学杂志

Designing a Forecast Model for Economic Growth of Japan Using Competitive (Hybrid ANN vs Multiple Regression) Models

Abstract

Ahmet Demir, Atabek Shadmanov, Cumhur Aydinli and Okan Eray

Artificial neural network models have been already used on many different fields successfully. However, many researches show that ANN models provide better optimum results than other competitive models in most of the researches. But does it provide optimum solutions in case ANN is proposed as hybrid model? The answer of this question is given in this research by using these models on modeling a forecast for GDP growth of Japan. Multiple regression models utilized as competitive models versus hybrid ANN (ANN + multiple regression models). Results have shown that hybrid model gives better responds than multiple regression models. However, variables, which were significantly affecting GDP growth, were determined and some of the variables, which were assumed to be affecting GDP growth of Japan, were eliminated statistically.