Predicting National Rice Production Using XGBoost Method Based Hyperparameter Optimization

The increasing population has an impact on increasing food needs, and increasing rice production is a top priority in human resource development. Indonesia has extensive agricultural land but still imports from other Southeast Asian countries, and the government has taken anticipation through surveys and observations at strategic points using technology. Therefore, this research aims to find a machine learning model to predict rice production in Indonesia with the smallest MAPE. The method used is XGBoost with Grid Search hyperparameters for time series forecasting models, Grid Search is used to find the best parameters to increase accuracy. Prediction results can be used as suggestions or to help make national policies to achieve food self-sufficiency. This research uses rice production data from the Ministry of Agriculture in the period 1970-2022. The results show that XGBoost performs well with a MAPE of 7.43% on training data. Moreover, when the Grid Search hyperparameter is used, it produces a MAPE of 6.73%, which increases the performance.