Comparisons of VAR Model and Models Created by Genetic Programming in Consumer Price Index Prediction in Vietnam

Abstract

In this paper, we present an application of GeneticProgramming (GP) to Vietnamese CPI inflation one-stepprediction problem. This is a new approach in buildinga good forecasting model, and then applying inflationforecasts in Vietnam in current stage. The studyintroduces the within-sample and the out-of-samplesone-step-ahead forecast errors which have positivecorrelation and approximate to a linear function withpositive slope in prediction models by GP. We alsobuild Vector Autoregression (VAR) model to forecast CPIin quarterly data and compare with the models createdby GP. The experimental results show that the GeneticProgramming can produce the prediction models havingbetter accuracy than Vector Autoregression models. Wehave no relevant variables (m2, ex) of monthly data inthe VAR model, so no prediction results exist tocompare with models created by GP and we just forecastCPI basing on models of GP with previous data of CPI.