Optimization of the Linear Regression Algorithm using GridSearchCV for Rice Crop Production Prediction

Abstract

Rice production in Central Java Province fluctuates annually, affecting food security and agricultural output distribution. Therefore, accurate prediction methods are essential to assist stakeholders in agricultural planning and strategic decision-making. This study applies the Linear Regression algorithm to predict rice production based on historical data from 2014 to 2023 obtained from the official website of the Central Java Provincial Agriculture and Plantation Office. The model is developed using multiple linear regression with variables including planted area, harvested area, and productivity. The novelty of this study lies in the structured application of hyperparameter tuning using GridSearchCV to optimize Linear Regression performance, as well as the integration of a preprocessing pipeline based on data distribution stabilization to improve accuracy and model generalization. The research process includes data collection, preprocessing, modeling, optimization, model evaluation, and deployment as a web-based application using Streamlit Cloud. GridSearchCV optimization results indicate a cross-validation accuracy of 98.26%, confirming the model’s strong predictive capability. Model evaluation shows an R² value of 0.9754, with MAE of 0.0957, MSE of 0.0307, and RMSE of 0.1753, indicating low prediction errors and stable model performance. The optimized model is implemented as a web application via Streamlit Cloud, enabling direct use by end-users. For future research, it is recommended to incorporate additional variables such as rainfall, temperature, and rice variety, or to compare performance with other algorithms such as Random Forest, Support Vector Regression, or Long Short-Term Memory (LSTM) to further enhance prediction accuracy.