Comparison of Xgboost, Lstm, and Neural Prophet Models for Red Cayenne Pepper Price Prediction in East Java

Abstract

The price of red cayenne pepper in Indonesia, particularly in East Java Province, frequently experiences significant fluctuations, affecting food inflation and economic stability. Accurate price forecasting is therefore essential to support decision-making in food supply chain management and price stabilization policies. This study compares the forecasting performance of three models, namely XGBoost, LSTM, and Neural Prophet, using historical price data from September 2024 to August 2025 obtained from the official Siskaperbapo website of the East Java Provincial Department of Industry and Trade (DISPERINDAG). A quantitative time series forecasting approach was employed, and model performance was evaluated using Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The results indicate that XGBoost achieved the highest prediction accuracy, with an RMSE of 0.52, MAE of 0.40, and MAPE of 1.42%, outperforming LSTM and Neural Prophet. The findings also show that model selection, dataset length, and training-test split proportion significantly influence forecasting performance, with longer datasets and larger training sets generally improving prediction accuracy. Overall, XGBoost proved to be the most accurate and stable model for forecasting red cayenne pepper prices, providing valuable support for AI-based food price prediction and agricultural policy decision-making.