Forecasting Monthly Rice Stock in the Philippines Using Time Series Models

Abstract

This study investigated the status of rice stocks in the Philippines by analyzing data obtained from the Philippines Statistics Authority (PSA) spanning from January 2000 to March 2023. The exponential smoothing and Box-Jenkins methods were used to build a forecasting model for the rice stock in the Philippines. The different models for each method were evaluated in the training dataset using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Akaike's Information Criterion (AIC). The Holt-Winters(A,A,A) model and the ARIMA (0,1,0)(1,1,1)₁₂ model were the candidate models for the exponential smoothing method and Box-Jenkin method, respectively. The two candidate models had close performance in the training stage. However, the Holt-Winters(A,A,A) had smaller forecasting errors in the testing set. Thus, the final forecasting model was Holt-Winters(A,A,A). The forecast from the model suggests that Philippine rice stock may be enough to supply the country if there is no increase in the demand for rice until March 2024.