Coffee is one of the leading commodities in Indonesia because it has good market opportunities both at Indonesia and overseas. Coffee commodities often experience price fluctuations as a result of an imbalance between coffee demand and supply. One of the efforts to anticipate price fluctuations is to do price forecasting. Many forecasting methods can be used, such as Fuzzy Time Series and Long Short-Term Memory which are used in this study to predict coffee prices on the next day. This study will use data on prices for North Sumatra Arabica coffee from January 2020 to August 2021 which were obtained from the official BAPPEBTI's website. In this study, the MAPE value generated in the Fuzzy Time Series Average-Based Interval is 0.016 and the smallest MAPE value in the Long Short-Term Memory method is obtained when the learning rate = 0.00001 with an initial weight value of 0.5 which means obtained MAPE value of 0.06. The MAPE value of both methods is below 10, so it can be said that both methods are categorized as very good. It can be said that in this study the Fuzzy Time Series Average-Based Interval method has better accuracy than Long Short-Term Memory in forecasting the price of Arabica coffee in North Sumatra.
Copyrights © 2022