Forecasting Rice Yield in Thailand Using Machine Learning Techniques

Abstract

Linear regression and random forest regression (RFR), two machine learning approaches, are used in this study’s quantitative analysis to predict rice yield in Thailand. The analysis utilizes data from 2022 and 2023, encompassing variables pertinent to rice production, including production year, maximum and minimum rainfall, soil characteristics, maximum and minimum temperatures, and rice output. The models were developed with the intention of predicting rice yield, and their performance was assessed using the Mean Absolute Error (MAE) and coefficient of determination (R²). Experimental results demonstrate that the Random Forest Regression model surpasses the Linear Regression model, attaining a R² value of 38.19%, beyond the 30.45% achieved by the Linear Regression model. Nonetheless, the evaluation results indicate that both models continue to have a degree of predictive error. The research indicates prospective enhancements in model construction, emphasizing parameter optimization, the selection of pertinent factors influencing rice yield, and the augmentation of both the quality and quantity of data utilized for model training. Expanding the data scope to encompass a longer time frame and incorporate a broader range of variables would enhance forecast accuracy and reliability. This would enhance planning and policy development in Thailand’s agriculture industry.