Modeling of red onion production in Central Java using hybrid ARIMA-ANFIS

Abstract

Red onion is one of the strategic horticulture commodities in Indonesia considering its function as the main ingredients of the basic ingredients of Indonesian food. For increasing production to supply national necessary, Central Java as the main center of red onion production should be able to predict the production of several periods to maintain the balance of national production. The purpose of this research is to get the best model to forecast the production of red onion in Central Java by ARIMA, ANFIS, and hybrid ARIMA-ANFIS method. The smallest RMSE and AIC values measure model accuracy. The results show that the best model for modeling red onion production in Central Java is obtained by hybrid ARIMA - ANFIS model which is a combination between SARIMA ([2], 1, [12]) and residual ARIMA using ANFIS model with input et,1, et,2 on the grid partition technique, gbell membership function, and membership number of 2 that produce RMSE 12033 and AIC 21.6634. While ARIMA model yield RMSE 13301,24 and AIC 21,89807 with violation of assumption. And the ANFIS model produces RMSE 14832 and AIC 22,0777. It shows that ARIMA-ANFIS hybrid method is better than ARIMA and ANFIS.