PERFORMANCE OF THE ACCURACY OF FORECASTING THE CONSUMER PRICE INDEX USING THE GARCH AND ANN METHODS

Abstract

The Consumer Price Index (CPI) is the most widely used indicator of the inflation rate. Then, the value of CPI in the future must be known to be the basis of the government's making appropriate and accurate policies. The CPI data used in this study was taken from the Central Statistics Agency (BPS) from January 2006 - to December 2021. The CPI data used has a data pattern containing symptoms of heteroskedasticity. To overcome the symptoms of heteroskedasticity, the author uses the GARCH and ANN methods to determine the value of CPI in the future. The GARCH method can overcome the symptoms of heteroskedasticity in the time series forecasting process, while ANN is an effective method in time series forecasting because of its high level of accuracy. In this study, mape error calculation results were obtained with the ARIMA model (4,2,2)~GARCH(1.1) of 3.19% or with an accuracy of 96.81%, and ANN using two hidden layers of 1.24% or with an accuracy of 98.76%. Thus, the results of this study show that the ANN method is the best method of forecasting Consumer Price Index (CPI) data.