Forecasting Rice Prices in Indonesia Using a Hybrid HWES-MLP Time Series Prediction Model

Supriadin, Supriadin
Haris, M. Al
Amri, Saeful
Abas, Hafiza
Fadugba, Sunday Emmanuel

JTAM (Jurnal Teori dan Aplikasi Matematika)
Vol 10, No 2 (2026): April

Rice is the main staple food for the majority of the Indonesian population. However, the fluctuation in rice prices and future uncertainty emphasize the importance of forecasting rice prices, thus requiring a forecasting model capable of providing accurate predictions. Various previous forecasting methods have been limited in capturing the combination of linear and non-linear patterns in rice price data, spurring the need for a more comprehensive hybrid approach. This research applies a quantitative approach by utilizing secondary data sourced from publications of the Central Statistics Agency (BPS) of Indonesia. This study aims to forecast rice prices in Indonesia using a hybrid approach combining Holt–Winters Exponential Smoothing (HWES) with Multilayer Perceptron (MLP). The hybrid model is designed to overcome the limitations of the Holt-Winters Exponential Smoothing method, which can only capture linear patterns such as trend and seasonality, by adding the Multilayer Perceptron method to capture non-linear patterns that cannot be handled by the linear approach. The dataset comprises monthly rice prices in Indonesia from January 2010 to December 2024, while the period of January–December 2025 is used as the prediction period. The data analysis process was carried out using the software R-Studio and Minitab, which provide a variety of features to support time series modeling. The results indicate that the most effective method for forecasting rice prices in Indonesia is the Hybrid Holt Winters Exponential Smoothing (α = 0.5; β = 0.3; γ = 0.3)-Multilayer Perceptron (12-12-1), which achieved the highest accuracy with a MSE of 9666.12, a RMSE of 310.9117, and a MAPE of 1.9949%. This finding indicates that the Hybrid HWES-MLP approach is highly capable of capturing rice price data patterns. Thus, this model holds significant potential to be utilized as a benchmark supporting government policy in maintaining rice price stability, market intervention, and optimizing the management of national rice reserves stock.