Indonesia’s Food Commodity Price Forecasting using Recurrent Neural Networks

Managing strategic commodities prices in the market is considered an important task since they have a significant contribution to the calculation of the inflation rate. Inflation management has a strong connection to the public’s economic activities and buying power. To aid this problem, it is necessary to find the best forecasting model that can predict commodities daily price. This paper aims to find the best prediction model between Recurrent Neural Network (RNN) variants, LSTM and GRU, in forecasting the daily price of three Indonesia’s strategies commodities: rice, broiler meat, and chicken egg. The result shows that the GRU model achieves higher accuracy in predicting the daily price of rice, broiler meat, and chicken egg, based on two evaluation metrics Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE). The GRU model also managed to finish the computational process faster than LSTM by$\sim$20 seconds.