<![CDATA[The price of red chili in Jepara Regency, Central Java, shows high and nonlinear fluctuations due to seasonal factors, weather, and supply dynamics. This instability affects food inflation as well as the income of farmers and traders. This study aims to build a prediction model for red chili prices in Jepara Regency using the Long Short-Term Memory (LSTM) method and compare its performance with the Autoregressive Integrated Moving Average (ARIMA) method. The data used are weekly red chili prices from January 2020 to December 2025, obtained from the official hargajateng.org website. The research steps include data preprocessing, normalization using Min-Max Scaling, forming time series data with a sliding window, training the LSTM model using the Adam optimizer, Mean Squared Error (MSE) as the loss function, a maximum of 200 epochs, batch size of 8, and applying EarlyStopping to get the best model. Performance evaluation was done using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The study results show that the LSTM model can follow chili price change patterns well, producing an RMSE of 16,707.02, an MAE of 12,348.54, and a MAPE of 19.31% on the testing data. Additionally, the 16-week forecasting results give an idea of the price trend in the upcoming period. Compared to the ARIMA model, which produced an RMSE of 20,992.15, an MAE of 13,867.26, and a MAPE of 22.70%, the LSTM model shows better prediction performance with lower error rates. The study indicates that the LSTM method is more effective in modeling nonlinear patterns in red chili price data compared to the ARIMA method.]]>
Copyrights © 2026
