Comparative Performance Evaluation of ARIMA, SARIMA, and LSTM for Daily Shallot Price Forecasting in Palembang City

Miranda, Fatreisya Ayu
Tania, Ken Ditha
Kurnia, Rizka Dhini

Electronic Journal of Education, Social Economics and Technology
Vol 6, No 2 (2025)

Shallots are a food commodity that often experiences price fluctuations and is one of the contributors to inflation in the city of Palembang. This study compares the ARIMA, SARIMA, and LSTM methods in predicting shallot prices using daily data start from January 2020 to October 2025. The Data of shallot price were obtained through the official website of Bank Indonesia. The stages of the study included data collection, pre-processing, visualization and decomposition, split data, modeling, and performance evaluation using the RMSE, MAE, and MAPE metrics. Model performance assessment reveals that ARIMA(1,1,1) method provided the most optimal performance with the lowest error value in comparison with the remaining two other methods, namely SARIMA and LSTM. The SARIMA(1,1,1)(2,1,1)7 model produced a slightly higher error rate, although its performance remains superior than LSTM model. The LSTM method produced the highest error in this study. These findings indicate that the pattern of shallot price data in Palembang tends to follow linear and seasonal trends that are not too complex, so that classical statistical approaches are still superior to deep learning models in capturing these data patterns. This research provides practical contributions as a decision-making support tool for the government and business actors in planning the distribution and stabilization of shallot prices in Palembang City.