Open AccessInflation modeling in Indonesia using hybrid autoregressive integrated moving average (ARIMA)-neural network (NN)
Author(s) -
Dinda Pratiwi,
Mustika Hadijati
Publication year - 2021
Publication title -
iop conference series. materials science and engineering
Language(s) - English
Resource type - Journals
eISSN - 1757-899X
pISSN - 1757-8981
DOI - 10.1088/1757-899x/1115/1/012058
Subject(s) - autoregressive integrated moving average , mean squared error , inflation (cosmology) , residual , econometrics , statistics , autoregressive model , mean absolute percentage error , box–jenkins , mathematics , artificial neural network , variance inflation factor , time series , linear regression , computer science , algorithm , artificial intelligence , multicollinearity , physics , theoretical physics
One of the macroeconomic indicators to see the stability of a country’s economy is inflation. This study aims to model the value of monthly inflation in Indonesia from January 2003 to December 2019 using the ARIMA-NN hybrid. The data plot shows a non-linear pattern and trends, so that the differencing process is carried out and the model is built using ARIMA model. The best ARIMA model obtained is SARIMA (1,1,0)(0,1,1) 12 with a Root Mean Square Error (RMSE) of 0.01134. Furthermore, ARIMA residuals that do not satisfy white noise and normality are modeled using NN. The best structure obtained of NN model is (3×2×1) with an RMSE of 0.023984. From the ARIMA and residual NN prediction results, the ARIMA-NN hybrid model is obtained to predict the value of monthly inflation in Indonesia for the next 12 months with the Mean Absolut Percentage Error (MAPE) value is 11.40873%. It means that the model result has high prediction accuracy.