Advancing Prediction of Rice Production in Sumatra: Utilizing eXtreme Gradient Boosting Algorithm for Accurate Forecasts

Rice production in Sumatra is crucial to food security in Indonesia. However, accurately predicting rice production is challenging due to climate change and agricultural land conversion. To address this challenge, this research proposes utilizing XGBoost with feature engineering to improve the accuracy of rice production prediction on Sumatra Island. This research aims to develop a prediction model capable of understanding and anticipating fluctuations in rice production by analyzing historical data encompassing rice production, land area, and climate from 1993 to 2020. The performance of the proposed prediction model was assessed using R-squared ($\mathrm{R}^{2}$), Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) as primary evaluation metrics. The results show that through feature engineering, the prediction model can achieve the highest accuracy, especially with a 90:10 data split resulting in an $R^{2}$ value of $\mathbf{0. 9 7 6}$, RMSE and MAE values of $\mathbf{0. 1 0 9}$ and 0.076. This result indicates that the application of feature engineering techniques and large training data considerably improves the accuracy of the prediction model.