Modeling the Relationship between Rice Yield and Climate Variables Using Statistical and Machine Learning Techniques

This paper presents the application of a multiple number statistical methods and machine learning techniques to model relationship between rice yield climate variables major region in Sri Lanka, which contributes significantly country’s paddy harvest. Rainfall, temperature (minimum maximum), evaporation, average wind speed (morning evening), sunshine hours are climatic factors considered for modeling. Rice harvest data over last three decades monthly were used develop prediction by applying artificial neural networks (ANNs), support vector regression (SVMR), linear (MLR), Gaussian process (GPR), power (PR), robust (RR). The performance each was assessed terms mean squared error (MSE), correlation coefficient (R), absolute percentage (MAPE), root ratio (RSR), BIAS value, Nash number, it found that GPR-based is most accurate among them. Climate collected until early 2019 (Maha season year 2018) model, an independent validation performed Yala 2019. developed can be forecast future with very high accuracy.