Indonesia White Sugar Supply and Demand Forecast Using Machine Learning

This paper aims to provide the best Machine Learning model to forecast Indonesia's white sugar supply and demand to provide potential supply chain efficiency and price stability. The high price of white sugar in Indonesia is due to low productivity, imports, and high demand. In midyear 2018, domestic sugar reached almost three times higher than the international market price, reflecting the commodity scarcity in the country. Previous research has been carried out to forecast sugar prices in Thailand using machine learning and consumption data in 4 years. The nonstationary data used in this research include 20 years of historical data in import and production as supply and demand data, research focused on this completeness and fulfilment of this data. Forecasting data using deep learning with LSTM is the best method model that produces the best performance compared with regressor models, which has not been done previously. The decrease in LSTM evaluation method compared to AdaboostRegressor was 37.33% for MSE and 14.12% for MAE. As a result, LSTM, as the best model, is confirmed by the most petite MAE and MSE values.