AI Meets Agriculture: Advanced Rice Yield Forecasting Based on Historical Data using Machine Learning Approaches

Despite the global importance of agriculture, inefficient practices persist, particularly in poorer regions. A big obstacle to increasing agricultural outputs is the inefficiency of conventional methods and the absence of accurate instruments for predicting crop yields. While various machine learning techniques have been used by several researchers to forecast yields, the difficulty lies in developing models that adequately account for regional nuances. To overcome this problem, this study proposed a novel machine learning-based approach for crop yield prediction focused on the Sylhet region, Bangladesh with a large unique dataset named "EcoRiceDynamics (1971-2021)." Four models—Linear Regression, Random Forest, Gradient Boosted, and K-Nearest Neighbors are evaluated to improve forecast precision. Performance was measured using the R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value and error reduction metrics. Our results show that the Gradient Boosted model achieved the highest R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 95% and MAE 6.71%, indicating its superior fit to the dataset. However, the Random Forest model excelled in minimizing prediction errors by 6.37% with R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> value of 94.1%. Random Forest reduces prediction errors effectively, while Gradient Boosted explains variance better. Incorporating these models into agricultural systems could enhance decision-making and crop yields in regions like Sylhet. This study highlights the potential of machine learning in advancing agricultural sustainability on yield estimation.