A SUPPORT VECTOR REGRESSION APPROACH FOR FORECASTING RICE PRODUCTION IN INDONESIA BASED ON AGRONOMIC AND CLIMATE FACTORS

Abstract

Rice production plays a critical role in ensuring national food security in Indonesia, where rice remains the primary staple food. However, rice yield stability is highly vulnerable to agronomic variability and climate dynamics, making accurate forecasting essential for strategic agricultural planning. This study proposes a rice production forecasting model using Support Vector Regression (SVR), a machine learning algorithm well known for its robustness in handling non-linear relationships and multivariate data. The dataset consists of provincial-level agricultural indicators—harvested area and productivity—combined with climate variables, including rainfall, temperature, and humidity, collected from open sources (BMKG and Kaggle) for the period 2018–2024. The research methodology includes data preprocessing (missing value imputation and Min–Max normalization), exploratory data analysis, model training, and performance evaluation using Mean Squared Error (MSE) and coefficient of determination (R²). Experimental results indicate that the SVR model achieved optimal performance using a 90:10 training–testing split, yielding an MSE of 0.0004 and an R² of 0.9918 on normalized data, demonstrating strong predictive capability. Harvested area and productivity emerged as dominant predictors, while climate variables contributed non-linear effects captured effectively by the SVR model. These findings suggest that SVR is a reliable and interpretable approach for rice yield forecasting and has strong potential as a decision-support tool for agricultural policy and food security management in Indonesia.