Early warning system modeling for rice supply using backpropagation artificial neural network to manage imported rice

Abstract

Rice is a staple food in Indonesia. Although Indonesia produces a large amount of rice, it cannot meet domestic rice needs. The unpredictable domestic rice supply prompted the government to import rice. Moreover, rice imports are one of the efforts to provide rice stock. On the other hand, importing rice can decrease domestic rice prices because it creates a market competitor. This study uses backpropagation artificial neural networks to develop a prediction system for rice supply crises in Indonesia based on models similar to currency crisis prediction systems. The study identified key variables and indicators for predicting rice supply crises, including rice production, consumption, prices, land area, and population. Data from January 2012 to December 2022 was analyzed. The optimal neural network architecture achieved a Mean Squared Error (MSE) of 0.209192. The analysis revealed that rice consumption, land area, and total population are the most strongly correlated indicators of a rice commodity crisis