Forecasting Palm Oil Production Using Long Short-Term Memory (LSTM) With Time Series Cross Validation (TSCV)

Abstract

Oil palm plant (Elaeis guineensis Jacq.) is a plantation crop that has a high economic value for Indonesia, because the results of oil palm plantations can increase the country's foreign exchange. Oil palm plantations can create jobs for the people of Indonesia, thus reducing unemployment in Indonesia. Oil palm plantations in Indonesia have spread to various regions, besides being found on the islands of Sumatra and Kalimantan, now oil palm plantations are almost found in various regions in Indonesia both small-scale plantations and large-scale plantations. This research uses historical data in the form of monthly palm oil production to predict the price of strategic food commodities. The period of palm oil production used is from January 1997 to December 2023 obtained from the website of PT. X or documentation data at PT. X. In this study the data was divided into 4 data scenarios using the Time Series Cross Validation (TSCV) method. The results of palm oil production modeling with LSTM that have been carried out show that palm oil production data shows differences in forecasting values and accuracy in the number of neurons and epochs used. The conclusion of this study is that from data processing and analysis in the previous chapter, it can be concluded that forecasting the amount of palm oil production in PT X can be modeled with LTSM and SARIMA method using Time Series Cross Validation (TSCV) data.