Forecasting Indonesian inflation using a hybrid ARIMA-ANFIS

Abstract

This paper discusses the prediction of the inflation rate in Indonesia. The data used in this research is assumed to have both linear and non-linear components. The ARIMA model is selected to accommodate the linear component, while the ANFIS method accounts for the non-linear component in the inflation data. Thus, the model is known as the hybrid ARIMA-ANFIS model. The clustering method is performed in the ANFIS model using Fuzzy C-Mean (FMS) with a Gaussian membership function. Consider 2 to 6 clusters. The optimal number of clusters is assessed according to the minimum value of the error prediction. To evaluate the performance of the fitted hybrid ARIMA-ANFIS model, it can be compared to the classical ARIMA model and with the ordinary ANFIS model. The result reveals that the best ARIMA model for inflation prediction in Indonesia is ARIMA(2,1,0). In the hybrid ARIMA(2,1,0)-ANFIS model, two clusters are optimal. Meanwhile, the optimum number of clusters in the ordinary ANFIS model is six. The comparison of prediction accuracy confirms that the hybrid model is superior to the individual model alone of either ARIMA or ANFIS model.