Evaluasi Kinerja Model Long Short-Term Memory Pada Prediksi Produksi Bawang Merah Kabupaten Brebes

Abstract

Shallot production is one of the most important horticultural commodities that contributes to national food security and regional economic development, particularly in Brebes Regency, one of the largest shallot-producing areas in Indonesia. Production fluctuations caused by weather conditions and agricultural factors necessitate the use of accurate forecasting methods. This study aims to evaluate the performance of the Long Short-Term Memory (LSTM) method in predicting shallot production in Brebes Regency. The dataset consisted of production, harvested area, rainfall, and temperature data from 2016 to 2025, comprising 1,288 records. The research process included data preprocessing, feature engineering, Min-Max Scaling normalization, LSTM model development, and performance evaluation using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the coefficient of determination (R²). The proposed model employed two LSTM layers with 64 and 32 neurons, a dropout rate of 0.2, and a dense layer with 16 neurons. Experimental results showed that the model achieved an MAE of 9,478.55 quintals, an RMSE of 15,763.86 quintals, and an R² value of 0.8990. These results indicate that the model can explain 89.90% of the variation in shallot production data. Therefore, the LSTM model demonstrates excellent predictive performance and has the potential to support decision-making in agricultural production planning in Brebes Regency