Forecasting Inflation with the Influence of Globalization using Artificial Neural Network-based Thin and Thick Models

We study globalization influences on forecasting inflation in an aggregate perspective using the Phillips curve for Hong Kong, Japan, Taiwan and the US by artificial neural network-based thin and thick models. Our empirical results support the hypothesis that globalization influences do generate a downward tendency in inflation through time in all cases with different levels. Moreover, the artificial neural network-based thin and thick models that have been developed upon the best linear model for each country have shown significant superiority over the naïve model in the most cases and over the best linear model in some cases. With the virtue of application flexibility, finding optimal values of all the parameters for so many artificial neural network models is very time-consuming and complex because of a large number of parameters considered. Thus we do not make any claim on the optimality of the artificial neural network models in this study. As a consequence, although our empirical results are moderately satisfactory, building artificial neural network models based upon the best linear model is a good compromise between practicality and optimality.