期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2021
卷号：99
期号：1
页码：75
出版社：Journal of Theoretical and Applied
摘要：This research was proposed for constructing the predictive model of commodity price data. The classical model such as Autoregressive Integrated Moving Average (ARIMA) and also machine learning model such Adaptive Neuro-Fuzzy Inference System (ANFIS) have been implemented in various field of time series analysis. This research is focused on constructing ARIMA, ANFIS and their combination or ensemble ARIMA-ANFIS. The main problem of combination is determining the weight of each vector predicted values which obtained from related models. In this research, the weight of each model were determined by variance-covariance approach and Lagrange Multiplier optimization, while in classical studies weight of each model was determined by averaging of predicted values. The main issue of this research is how to determine the weight of vector predicted values by using variance-covariance approach for constructing the ensemble ARIMA-ANFIS. The daily data of garlic price in Semarang collected from January 2019 to August 2019 were used as case studies. ARIMA, ANFIS and ensemble ARIMA-ANFIS were implemented for predicting data. ARIMA individual, ANFIS individual, ensemble model by averaging and ensemble model by weighting resulted high accuracy for predicting. The combination of ARIMA(1,0,0)-ARCH(1) and ANFIS (with lag-1, lag-2 as inputs and 2 MFs) is the best model for forecasting garlic price data in Semarang. The MAPE values of all models were less than 5% which had shown a good performance for forecasting.
